You searched for +publisher:"University of Michigan" +contributor:("Lewis, Richard L").
Showing records 1 – 30 of
36 total matches.
◁ [1] [2] ▶
University of Michigan
1.
Shim, Jae-Young.
The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions.
Degree: PhD, Linguistics, 2015, University of Michigan
URL: http://hdl.handle.net/2027.42/116623 
► Prevailing analyses of Noun Phrases with a topic or focus interpretation in the literature of the Minimalist Program depend heavily on a syntactic projection (e.g.,…
(more)
▼ Prevailing analyses of Noun Phrases with a topic or focus interpretation in the literature of the Minimalist Program depend heavily on a syntactic projection (e.g., Topic Phrase) or a syntactic feature (e.g., [+foc]) dedicated to Topic or Focus. For example, Rizzi (1997) utilizes both a Topic Phrase and a Focus Phrase to account for a topic and focus interpretation of Noun Phrases. Miyagawa (2010), another influential work on Topic and Focus within the Minimalist Program, also employs the notion Focus as a syntactic feature that operates in the syntax.
This dissertation, contrary to the prevailing analyses briefly described above, aims to present an alternative account of Noun Phrases with a topic or focus interpretation. In the account, I argue, without appealing to a syntactic projection or feature dedicated to Topic or Focus, that a topic or focus interpretation of Noun Phrases is calculated and determined at the Conceptual-Intentional interface by means of the structural properties of Noun Phrases.
To achieve this goal, this dissertation explores Topic- and Focus-related interpretations of Noun Phrases in typical transitive constructions in Korean and proposes that the presence of a contrastively marked Noun Phrase, whether it is Topic or Focus, plays a crucial role in determining a topic or focus interpretation of Noun Phrases in Korean.
Advisors/Committee Members: Epstein, Samuel D (committee member), Lewis, Richard L (committee member), Pires, Acrisio (committee member), Baptista, Marlyse (committee member).
Subjects/Keywords: Topic and Focus; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Shim, J. (2015). The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/116623
Chicago Manual of Style (16th Edition):
Shim, Jae-Young. “The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions.” 2015. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/116623.
MLA Handbook (7th Edition):
Shim, Jae-Young. “The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions.” 2015. Web. 04 Mar 2021.
Vancouver:
Shim J. The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions. [Internet] [Doctoral dissertation]. University of Michigan; 2015. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/116623.
Council of Science Editors:
Shim J. The Contrast-dependent CI-Calculation of Topic and Focus in Korean Transitive Constructions. [Doctoral Dissertation]. University of Michigan; 2015. Available from: http://hdl.handle.net/2027.42/116623
2.
De Los Santos, Guadalupe.
Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners.
Degree: PhD, Psychology, 2017, University of Michigan
URL: http://hdl.handle.net/2027.42/138538
► Bilingualism is prevalent, with over half of the population of the world being bilingual. While bilinguals have traditionally been viewed as having two separate languages,…
(more)
▼ Bilingualism is prevalent, with over half of the population of the world being bilingual. While bilinguals have traditionally been viewed as having two separate languages, modern views of language suggest that languages are not completely separate in the mind. This is especially evident in cases of intrasentential code-switching, when a speaker switches languages mid-sentence. Such points are of interest because they represent cases when the languages are activated simultaneously.
This dissertation expands our understanding of multi-language representation by investigating whether some grammatical representations generated by Spanish-English bilinguals and Spanish L2 language learners during reading are specific to the input language used to create the representation, or whether those representations are language-independent. Using eye-tracking, we measured reading times on nouns in grammatical (determiner-noun) and ungrammatical (adverb-noun) contexts, in both same language and mixed language pairs, as participants performed a two-string lexical decision task.
Experiment 1 found that bilinguals read nouns faster following determiners than adverbs. Crucially, this grammatical predictability effect did not interact with the same/mixed language variable. This suggests that grammatical predictability in this context is language-independent, not affected by language nor the presence of a language switch. Experiment 2 found a similar pattern for Spanish language learners, though it was not significant. When the data for Experiment 1 and Experiment 2 were combined, there was a main effect of grammaticality that did not interact with language congruency, suggesting that language-independent predictions influenced reading times for both bilinguals and language learners.
Experiment 3 took into account categorical ambiguity, i.e., that the same word can belong to more than one grammatical class. We computed two conditional probabilities over abstract grammatical categories to represent grammaticality in a more fine-grained way, allowing syntactic category ambiguity. Participants read the second word faster as its probability given the category of the first word increased. This grammatical predictability effect was language-independent, in that it was not modulated by a language switch.
Overall, this dissertation provides an in-depth investigation into multi-language representation and grammatical predictability in Spanish/English bilinguals, focusing on syntactic sequences that have the same word order in the two languages. Our results most strongly support the shared syntax view of bilingual language representation, having found language-independent grammatical predictability across experiments.
Advisors/Committee Members: Boland, Julie E (committee member), Brennan, Jonathan R (committee member), Ellis, Nick (committee member), Lewis, Richard L (committee member).
Subjects/Keywords: psycholinguistics; bilingualism; eyetracking; bilingual syntactic representation; code-switching; Psychology; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
De Los Santos, G. (2017). Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/138538
Chicago Manual of Style (16th Edition):
De Los Santos, Guadalupe. “Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners.” 2017. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/138538.
MLA Handbook (7th Edition):
De Los Santos, Guadalupe. “Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners.” 2017. Web. 04 Mar 2021.
Vancouver:
De Los Santos G. Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners. [Internet] [Doctoral dissertation]. University of Michigan; 2017. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/138538.
Council of Science Editors:
De Los Santos G. Using Eye-tracking to Examine Grammatical Predictability in Spanish-English Bilinguals and Spanish Language Learners. [Doctoral Dissertation]. University of Michigan; 2017. Available from: http://hdl.handle.net/2027.42/138538
University of Michigan
3.
Ke, Hezao.
The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions.
Degree: PhD, Linguistics, 2019, University of Michigan
URL: http://hdl.handle.net/2027.42/151472
► The overall goal of this dissertation is to establish a linking theory between the syntax and semantics and the processing of subject-verb agreement and reflexive…
(more)
▼ The overall goal of this dissertation is to establish a linking theory between the syntax and semantics and the processing of subject-verb agreement and reflexive binding. This dissertation develops a unified syntactic analysis of agreement, based on a formalization of minimal search. Such an analysis accounts for a variety of agreement patterns observed in various languages, including negative concord in Czech, inflection doubling in Norwegian, Frisian and Swedish, multiple agree in Japanese, cyclic agree in Georgian and Hindi Urdu, and subject-complementizer agreement in Lubukusu. The minimal search-based analysis is also extended to reflexive binding. The minimal search-based analysis of subject-verb agreement and reflexive binding captures the syntactic similarity between these two constructions.
This dissertation then argues that subject-verb agreement and reflexive binding have an important representational difference: the phi-features involved in subject-verb agreement and reflexive binding are essentially different in their semantic content. The phi-features on bound reflexives (and bound variables generally) have semantic content and are semantically interpretable, whereas those on agreeing verbs/T heads are not semantically interpretable.
The syntactic and semantic analyses of agreement and reflexive binding have crucial consequences for the sentence processing study of subject-verb agreement and reflexive binding. This dissertation proposes that in cue-based retrieval, mismatches of semantically interpretable retrieval cues (in reflexive binding) are less tolerable to the parser than mismatches of phonological cues (in subject-verb agreement). Experimental results are provided to evaluate this Asymmetry of Interpretability Hypothesis. The results reveal that the (un)acceptability of a target sentence by an experimental participant influences the occurrence of illusions of grammaticality/facilitatory effects for that participant: facilitatory effects do not occur for grammatical or acceptable sentences. The results also indicate that there are significant contributing factors to the emergence of illusions of grammaticality/facilitatory effects, involving relative differences between the target and distractor with respect to frequency, phonological length, orthographic length, phonological neighborhood density, and orthographic neighborhood density. These results suggest that the distractors are not completely ignored in the retrieval of the relevant target in subject-agreement and reflexive binding.
Advisors/Committee Members: Epstein, Samuel D (committee member), Pires, Acrisio M (committee member), Lewis, Richard L (committee member), Brennan, Jonathan R (committee member).
Subjects/Keywords: subject-verb agreement; reflexive binding; illusions of grammaticality; cue-based retrieval; minimal search; phi-features; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Ke, H. (2019). The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/151472
Chicago Manual of Style (16th Edition):
Ke, Hezao. “The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions.” 2019. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/151472.
MLA Handbook (7th Edition):
Ke, Hezao. “The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions.” 2019. Web. 04 Mar 2021.
Vancouver:
Ke H. The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions. [Internet] [Doctoral dissertation]. University of Michigan; 2019. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/151472.
Council of Science Editors:
Ke H. The Syntax, Semantics and Processing of Agreement and Binding Grammatical Illusions. [Doctoral Dissertation]. University of Michigan; 2019. Available from: http://hdl.handle.net/2027.42/151472
University of Michigan
4.
Zhang, Shun.
Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety.
Degree: PhD, Computer Science & Engineering, 2020, University of Michigan
URL: http://hdl.handle.net/2027.42/163125
► When a computational agent (called the “robot”) takes actions on behalf of a human user, it may be uncertain about the human’s preferences. The human…
(more)
▼ When a computational agent (called the “robot”) takes actions on behalf of a human user, it may be uncertain about the human’s preferences. The human may initially specify her preferences incompletely or inaccurately. In this case, the robot’s performance may be unsatisfactory or even cause negative side effects to the environment. There are approaches in the literature that may solve this problem. For example, the human can provide some demonstrations which clarify the robot’s uncertainty. The human may give real-time feedback to the robot’s behavior, or monitor the robot and stop the robot when it may perform anything dangerous. However, these methods typically require much of the human’s attention. Alternatively, the robot may estimate the human’s true preferences using the specified preferences, but this is error-prone and requires making assumptions on how the human specifies her preferences.
In this thesis, I consider a querying approach. Before taking any actions, the robot has a chance to query the human about her preferences. For example, the robot may query the human about which trajectory in a set of trajectories she likes the most, or whether the human cares about some side effects to the domain. After the human responds to the query, the robot expects to improve its performance and/or guarantee that its behavior is considered safe by the human.
If we do not impose any constraint on the number of queries the robot can pose, the robot may keep posing queries until it is absolutely certain about the human’s preferences. This may consume too much of the human’s cognitive load. The information obtained in the responses to some of the queries may only marginally improve the robot’s performance, which is not worth the human’s attention at all. So in the problems considered in this thesis, I constrain the number of queries that the robot can pose, or associate each query with a cost. The research question is how to efficiently find the most useful query under such constraints.
Finding a provably optimal query can be challenging since it is usually a combinatorial optimization problem. In this thesis, I contribute to providing efficient query selection algorithms under uncertainty. I first formulate the robot’s uncertainty as reward uncertainty and safety-constraint uncertainty. Under only reward uncertainty, I provide a query selection algorithm that finds approximately-optimal k-response queries. Under only safety-constraint uncertainty, I provide a query selection algorithm that finds an optimal k-element query to improve a known safe policy, and an algorithm that uses a set-cover-based query selection strategy to find an initial safe policy. Under both types of uncertainty simultaneously, I provide a batch-query-based querying method that empirically outperforms other baseline querying methods.
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Durfee, Edmund H (committee member), Lewis, Richard L (committee member), Lasecki, Walter (committee member).
Subjects/Keywords: human-agent interaction; Markov decisions processes; query selection algorithms; planning under uncertainty; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Zhang, S. (2020). Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/163125
Chicago Manual of Style (16th Edition):
Zhang, Shun. “Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety.” 2020. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/163125.
MLA Handbook (7th Edition):
Zhang, Shun. “Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety.” 2020. Web. 04 Mar 2021.
Vancouver:
Zhang S. Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety. [Internet] [Doctoral dissertation]. University of Michigan; 2020. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/163125.
Council of Science Editors:
Zhang S. Efficiently Finding Approximately-Optimal Queries for Improving Policies and Guaranteeing Safety. [Doctoral Dissertation]. University of Michigan; 2020. Available from: http://hdl.handle.net/2027.42/163125
University of Michigan
5.
Zhang, Qi.
Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination.
Degree: PhD, Computer Science & Engineering, 2020, University of Michigan
URL: http://hdl.handle.net/2027.42/162948
► In a large number of real world domains, such as the control of autonomous vehicles, team sports, medical diagnosis and treatment, and many others, multiple…
(more)
▼ In a large number of real world domains, such as the control of autonomous vehicles, team sports, medical diagnosis and treatment, and many others, multiple autonomous agents need to take actions based on local observations, and are interdependent in the sense that they rely on each other to accomplish tasks. Thus, achieving desired outcomes in these domains requires interagent coordination. The form of coordination this thesis focuses on is commitments, where an agent, referred to as the commitment provider, specifies guarantees about its behavior to another, referred to as the commitment recipient, so that the recipient can plan and execute accordingly without taking into account the details of the provider's behavior. This thesis grounds the concept of commitments into decision-theoretic settings where the provider's guarantees might have to be probabilistic when its actions have stochastic outcomes and it expects to reduce its uncertainty about the environment during execution.
More concretely, this thesis presents a set of contributions that address three core issues for commitment-based coordination: probabilistic commitment adherence, interpretation, and formulation. The first contribution is a principled semantics for the provider to exercise maximal autonomy that responds to evolving knowledge about the environment without violating its probabilistic commitment, along with a family of algorithms for the provider to construct policies that provably respect the semantics and make explicit tradeoffs between computation cost and plan quality. The second contribution consists of theoretical analyses and empirical studies that improve our understanding of the recipient's interpretation of the partial information specified in a probabilistic commitment; the thesis shows that it is inherently easier for the recipient to robustly model a probabilistic commitment where the provider promises to enable preconditions that the recipient requires than where the provider instead promises to avoid changing already-enabled preconditions. The third contribution focuses on the problem of formulating probabilistic commitments for the fully cooperative provider and recipient; the thesis proves structural properties of the agents' values as functions of the parameters of the commitment specification that can be exploited to achieve orders of magnitude less computation for 1) formulating optimal commitments in a centralized manner, and 2) formulating (approximately) optimal queries that induce (approximately) optimal commitments for the decentralized setting in which information relevant to optimization is distributed among the agents.
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Durfee, Edmund H (committee member), Lewis, Richard L (committee member), Sinha, Arunesh (committee member).
Subjects/Keywords: Multiagent Coordination; Sequential Decision Making; Commitment; Markov Decision Process; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Zhang, Q. (2020). Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/162948
Chicago Manual of Style (16th Edition):
Zhang, Qi. “Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination.” 2020. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/162948.
MLA Handbook (7th Edition):
Zhang, Qi. “Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination.” 2020. Web. 04 Mar 2021.
Vancouver:
Zhang Q. Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination. [Internet] [Doctoral dissertation]. University of Michigan; 2020. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/162948.
Council of Science Editors:
Zhang Q. Making and Keeping Probabilistic Commitments for Trustworthy Multiagent Coordination. [Doctoral Dissertation]. University of Michigan; 2020. Available from: http://hdl.handle.net/2027.42/162948
University of Michigan
6.
Sorg, Jonathan Daniel.
The Optimal Reward Problem: Designing Effective Reward for Bounded Agents.
Degree: PhD, Computer Science & Engineering, 2011, University of Michigan
URL: http://hdl.handle.net/2027.42/89705
► In the field of reinforcement learning, agent designers build agents which seek to maximize reward. In standard practice, one reward function serves two purposes. It…
(more)
▼ In the field of reinforcement learning, agent designers build agents which seek to maximize reward. In standard practice, one reward function serves two purposes. It is used to evaluate the agent and is used to directly guide agent behavior in the agent's learning algorithm.
This dissertation makes four main contributions to the theory and practice of reward function design. The first is a demonstration that if an agent is bounded – if it is limited in its ability to maximize expected reward – the designer may benefit by considering two reward functions. A designer reward function is used to evaluate the agent, while a separate agent reward function is used to guide agent behavior. The designer can then solve the Optimal Reward Problem (ORP): choose the agent reward function which leads to the greatest expected reward for the designer.
The second contribution is the demonstration through examples that good reward functions are chosen by assessing an agent's limitations and how they interact with the environment. An agent which maintains knowledge of the environment in the form of a Bayesian posterior distribution, but lacks adequate planning resources, can be given a reward proportional to the variance of the posterior, resulting in provably efficient exploration. An agent with poor modeling assumptions can be punished for visiting the areas of the state space it has trouble modeling, resulting in better performance.
The third contribution is the Policy Gradient for Reward Design (PGRD) algorithm, a convergent gradient ascent algorithm for learning good reward functions. Experiments in multiple environments demonstrate that using PGRD for reward optimization yields better agents than using the designer's reward directly as the agent's reward. It also outperforms the use of an evaluation function at the leaf-states of the planning tree.
Finally, this dissertation shows that the ORP differs from the popular work on potential-based reward shaping. Shaping rewards are constrained by properties of the environment and the designer's reward function, but they generally are defined irrespective of properties of the agent. The best shaping reward functions are suboptimal for some agents and environments.
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Lewis, Richard L. (committee member), Laird, John E. (committee member), Polk, Thad A. (committee member), Wellman, Michael P. (committee member).
Subjects/Keywords: Reinforcement Learning; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Sorg, J. D. (2011). The Optimal Reward Problem: Designing Effective Reward for Bounded Agents. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/89705
Chicago Manual of Style (16th Edition):
Sorg, Jonathan Daniel. “The Optimal Reward Problem: Designing Effective Reward for Bounded Agents.” 2011. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/89705.
MLA Handbook (7th Edition):
Sorg, Jonathan Daniel. “The Optimal Reward Problem: Designing Effective Reward for Bounded Agents.” 2011. Web. 04 Mar 2021.
Vancouver:
Sorg JD. The Optimal Reward Problem: Designing Effective Reward for Bounded Agents. [Internet] [Doctoral dissertation]. University of Michigan; 2011. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/89705.
Council of Science Editors:
Sorg JD. The Optimal Reward Problem: Designing Effective Reward for Bounded Agents. [Doctoral Dissertation]. University of Michigan; 2011. Available from: http://hdl.handle.net/2027.42/89705
7.
Medeiros, David J.
Formal Approaches to the Syntax and Semantics of Imperatives.
Degree: PhD, Linguistics, 2013, University of Michigan
URL: http://hdl.handle.net/2027.42/99962
► This dissertation examines imperatives, defined as a morphological class, offering a set of interrelated, formal hypotheses regarding their syntactic and semantic analysis. Three major questions…
(more)
▼ This dissertation examines imperatives, defined as a morphological class, offering a set of interrelated, formal hypotheses regarding their syntactic and semantic analysis. Three major questions guide this research. First, the distribution of imperatives is addressed with respect to the possibility of occurrence in syntactically embedded clauses, a context often believed to exclude imperatives across languages.
It is shown that morphological imperatives productively occur in embedded clauses in languages with rich (person) imperative morphology. The independently motivated theory of Feature Transfer (Chomsky 2008) explains the cross-linguistic variation, in that imperatives with rich morphology are treated as finite verbs, while imperatives in e.g. English are selected by an operator with interpretable 2nd person features, following aspects of the analysis for English-type imperatives in Zanuttini (2008) and Han (2000).
The semantic properties of morphological imperatives constitutes a second research area. Embedded (non-performative) imperatives as well as non-command imperatives in English motivate an analysis in which imperatives have the semantic value of a weak necessity modal. The operator which obligatorily selects imperatives in English-type languages and optionally selects imperatives in rich-morphology languages has as its semantic value a set of presuppositions (following Kaufmann 2012) which encode performativity. Imperatives in English therefore have two semantic components, a modal component (represented clause-internally) and a performative component (represented in the left-periphery).
Finally, I address suppletive imperatives and the hypothesis (Portner 2004) that these belong to a clause type which includes morphological imperatives. While I reject the clause-type hypothesis for imperatives, I argue that performative imperatives and suppletive imperatives share one aspect of meaning, namely the performative component. Suppletive imperatives and true imperatives can differ in terms of modal component, deriving the differences between suppletives and imperatives in terms of the fine-grained interpretive possibilities discussed by von Fintel & Iatridou (2010).
These hypotheses lead to the conclusion that imperatives do not have proto-typical semantic functions such as Ordering (Kaufmann 2012) or Requiring (Portner 2004). Instead, imperatives encode weak necessity modality and can (in some languages must) occur in performative contexts, with a syntactic distribution which is wider than previously thought.
Advisors/Committee Members: Pires, Acrisio (committee member), Epstein, Samuel D. (committee member), Lewis, Richard L. (committee member), Keshet, Ezra Russell (committee member).
Subjects/Keywords: Imperatives; Syntax; Semantics; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Medeiros, D. J. (2013). Formal Approaches to the Syntax and Semantics of Imperatives. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/99962
Chicago Manual of Style (16th Edition):
Medeiros, David J. “Formal Approaches to the Syntax and Semantics of Imperatives.” 2013. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/99962.
MLA Handbook (7th Edition):
Medeiros, David J. “Formal Approaches to the Syntax and Semantics of Imperatives.” 2013. Web. 04 Mar 2021.
Vancouver:
Medeiros DJ. Formal Approaches to the Syntax and Semantics of Imperatives. [Internet] [Doctoral dissertation]. University of Michigan; 2013. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/99962.
Council of Science Editors:
Medeiros DJ. Formal Approaches to the Syntax and Semantics of Imperatives. [Doctoral Dissertation]. University of Michigan; 2013. Available from: http://hdl.handle.net/2027.42/99962
8.
Bloch, Mitchell.
Computationally Efficient Relational Reinforcement Learning.
Degree: PhD, Computer Science & Engineering, 2018, University of Michigan
URL: http://hdl.handle.net/2027.42/145859
► Relational Reinforcement Learning (RRL) is a technique that enables Reinforcement Learning (RL) agents to generalize from their experience, allowing them to learn over large or…
(more)
▼ Relational Reinforcement Learning (RRL) is a technique that enables Reinforcement Learning (RL) agents to generalize from their experience, allowing them to learn over large or potentially infinite state spaces, to learn context sensitive behaviors, and to learn to solve variable goals and to transfer knowledge between similar situations. Prior RRL architectures are not sufficiently computationally efficient to see use outside of small, niche roles within larger Artificial Intelligence (AI) architectures. I present a novel online, incremental RRL architecture and an implementation that is orders of magnitude faster than its predecessors. The first aspect of this architecture that I explore is a computationally efficient implementation of an adaptive Hierarchical Tile Coding (aHTC), a kind of Adaptive Tile Coding (ATC) in which more general tiles which cover larger portions of the state-action space are kept as ones that cover smaller portions of the state-action space are introduced, using k-dimensional tries (k-d tries) to implement the value function for non-relational Temporal Difference (TD) methods. In order to achieve comparable performance for RRL, I implement the Rete algorithm to replace my k-d tries due to its efficient handling of both the variable binding problem and variable numbers of actions. Tying aHTCs and Rete together, I present a rule grammar that both maps aHTCs onto Rete and allows the architecture to automatically extract relational features in order to support adaptation of the value function over time. I experiment with several refinement criteria and additional functionality with which my agents attempt to determine if rerefinement using different features might allow them to better learn a near optimal policy. I present optimal results using a value criterion for several variants of BlocksWorld. I provide transfer results for BlocksWorld and a scalable Taxicab domain. I additionally introduce a Higher Order Grammar (HOG) that grants online, incremental RRL agents additional flexibility to introduce additional variables and corresponding relations as needed in order to learn effective value functions. I evaluate agents that use the HOG on a version of Blocks World and on an Adventure task. In summary, I present a new online, incremental RRL architecture, a grammar to map aHTCs onto the Rete, and an implementation that is orders of magnitude faster than its predecessors.
Advisors/Committee Members: Laird, John E (committee member), Lewis, Richard L (committee member), Baveja, Satinder Singh (committee member), Durfee, Edmund H (committee member).
Subjects/Keywords: Relational Reinforcement Learning; Rete; Adaptive Tile Coding; Online Learning; Sequential Decision Making; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Bloch, M. (2018). Computationally Efficient Relational Reinforcement Learning. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/145859
Chicago Manual of Style (16th Edition):
Bloch, Mitchell. “Computationally Efficient Relational Reinforcement Learning.” 2018. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/145859.
MLA Handbook (7th Edition):
Bloch, Mitchell. “Computationally Efficient Relational Reinforcement Learning.” 2018. Web. 04 Mar 2021.
Vancouver:
Bloch M. Computationally Efficient Relational Reinforcement Learning. [Internet] [Doctoral dissertation]. University of Michigan; 2018. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/145859.
Council of Science Editors:
Bloch M. Computationally Efficient Relational Reinforcement Learning. [Doctoral Dissertation]. University of Michigan; 2018. Available from: http://hdl.handle.net/2027.42/145859
9.
Oh, Junhyuk.
Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration.
Degree: PhD, Computer Science & Engineering, 2018, University of Michigan
URL: http://hdl.handle.net/2027.42/145829
► Reinforcement learning (RL) is a general-purpose machine learning framework, which considers an agent that makes sequential decisions in an environment to maximize its reward. Deep…
(more)
▼ Reinforcement learning (RL) is a general-purpose machine learning framework, which considers an agent that makes sequential decisions in an environment to maximize its reward. Deep reinforcement learning (DRL) approaches use deep neural networks as non-linear function approximators that parameterize policies or value functions directly from raw observations in RL.
Although DRL approaches have been shown to be successful on many challenging RL benchmarks, much of the prior work has mainly focused on learning a single task in a model-free setting, which is often sample-inefficient. On the other hand, humans have abilities to acquire knowledge by learning a model of the world in an unsupervised fashion, use such knowledge to plan ahead for decision making, transfer knowledge between many tasks, and generalize to previously unseen circumstances from the pre-learned knowledge. Developing such abilities are some of the fundamental challenges for building RL agents that can learn as efficiently as humans.
As a step towards developing the aforementioned capabilities in RL, this thesis develops new DRL techniques to address three important challenges in RL: 1) planning via prediction, 2) rapidly generalizing to new environments and tasks, and 3) efficient exploration in complex environments.
The first part of the thesis discusses how to learn a dynamics model of the environment using deep neural networks and how to use such a model for planning in complex domains where observations are high-dimensional. Specifically, we present neural network architectures for action-conditional video prediction and demonstrate improved exploration in RL. In addition, we present a neural network architecture that performs lookahead planning by predicting the future only in terms of rewards and values without predicting observations. We then discuss why this approach is beneficial compared to conventional model-based planning approaches.
The second part of the thesis considers generalization to unseen environments and tasks. We first introduce a set of cognitive tasks in a 3D environment and present memory-based DRL architectures that generalize better to previously unseen 3D environments compared to existing baselines. In addition, we introduce a new multi-task RL problem where the agent should learn to execute different tasks depending on given instructions and generalize to new instructions in a zero-shot fashion. We present a new hierarchical DRL architecture that learns to generalize over previously unseen task descriptions with minimal prior knowledge.
The third part of the thesis discusses how exploiting past experiences can indirectly drive deep exploration and improve sample-efficiency. In particular, we propose a new off-policy learning algorithm, called self-imitation learning, which learns a policy to reproduce past good experiences. We empirically show that self-imitation learning indirectly encourages the agent to explore reasonably good state spaces and thus significantly improves sample-efficiency on RL domains where…
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Lee, Honglak (committee member), Lewis, Richard L (committee member), Deng, Jia (committee member).
Subjects/Keywords: Deep Reinforcement Learning; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Oh, J. (2018). Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/145829
Chicago Manual of Style (16th Edition):
Oh, Junhyuk. “Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration.” 2018. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/145829.
MLA Handbook (7th Edition):
Oh, Junhyuk. “Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration.” 2018. Web. 04 Mar 2021.
Vancouver:
Oh J. Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration. [Internet] [Doctoral dissertation]. University of Michigan; 2018. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/145829.
Council of Science Editors:
Oh J. Efficient Deep Reinforcement Learning via Planning, Generalization, and Improved Exploration. [Doctoral Dissertation]. University of Michigan; 2018. Available from: http://hdl.handle.net/2027.42/145829
10.
Cohn, Robert W.
Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis.
Degree: PhD, Computer Science and Engineering, 2016, University of Michigan
URL: http://hdl.handle.net/2027.42/120772
► An agent acting under uncertainty regarding how it should complete the task assigned to it by its human user can learn more about how it…
(more)
▼ An agent acting under uncertainty regarding how it should complete the task assigned to it by its human user can learn more about how it should behave by posing queries to its human user. Asking too many queries, however, may risk requiring undue attentional demand of the user, and so the agent should prioritize asking the most valuable queries. For decision-making agents, Expected Value of Information (EVOI) measures the value of a query, and so given a set of queries the agent can ask, the agent should ask the query that is expected to maximally improve its performance by selecting the query with highest EVOI in its set. Unfortunately, to compute the EVOI of a query, the agent must consider how each possible response would influence its future behavior, which makes query selection particularly challenging in settings where planning the agent's behavior would be expensive even without the added complication of considering queries to ask, especially when there are many potential queries the agent should consider. The focus of this dissertation is on developing query selection algorithms that can be feasibly applied to such settings.
The main novel approach studied, Wishful Query Projection (WQP), is based on the intuition that the agent should consider which query to ask on the basis of obtaining knowledge that would help it resolve a particular dilemma that it wishes could be resolved, as opposed to blindly searching its entire query set in hopes of finding one that would give it valuable knowledge. In implementing WQP, this dissertation contributes algorithms that are founded upon the following novel result: for myopic settings, when the agent can ask any query as long as the query has no more than some set number of possible responses, the best query takes the form of asking the user to choose from a specified subset of ways for the agent to behave. The work presented shows that WQP selects queries with near-optimal EVOI when the agent's query set is (1) balanced in the range of queries it contains; and (2) rich in terms of the highest contained query EVOI.
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Durfee, Edmund H (committee member), Lewis, Richard L (committee member), Wellman, Michael P (committee member).
Subjects/Keywords: Expected Value of Information; Query Selection Algorithms; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Cohn, R. W. (2016). Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/120772
Chicago Manual of Style (16th Edition):
Cohn, Robert W. “Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis.” 2016. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/120772.
MLA Handbook (7th Edition):
Cohn, Robert W. “Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis.” 2016. Web. 04 Mar 2021.
Vancouver:
Cohn RW. Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis. [Internet] [Doctoral dissertation]. University of Michigan; 2016. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/120772.
Council of Science Editors:
Cohn RW. Maximizing Expected Value of Information in Decision Problems by Querying on a Wish-to-Know Basis. [Doctoral Dissertation]. University of Michigan; 2016. Available from: http://hdl.handle.net/2027.42/120772
11.
Bartek, Brian D.
Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context.
Degree: PhD, Psychology, 2011, University of Michigan
URL: http://hdl.handle.net/2027.42/86317
► Understanding how short-term memory shapes sentence comprehension processes is a long-standing topic in psycholinguistics. This thesis pursues new insights on two facets of short-term memory’s…
(more)
▼ Understanding how short-term memory shapes sentence comprehension processes is a long-standing topic in psycholinguistics. This thesis pursues new insights on two facets of short-term memory’s role in sentence comprehension: (a) The first four experiments search for, and obtain, concrete evidence that locality effects, or increased integration difficulty attending increased dependent-head distance, occur even in simple sentence, where memory-based theories predict them and other theo- ries (such as experience- or expectation-based theories) do not. (b) The remainder of the thesis investigates the functional interplay between similarity-based interference and expectation. Expectation has been argued to facilitate processing at verbs by pre-activating lexical representations, and similarity between representations of recently linguistic context has been shown to slow processing of verbs by slowing integration of its dependencies with arguments; yet the relationship between these two potentially interacting processes has received neither theoretical nor empirical attention. This thesis presents a novel eyetracking experiment that finds effects of both expectation-based facilitation and retrieval-based difficulty on the integration of a subject-verb dependency, but no evidence of an interaction. This evidence sup- ports a simple model of expectation and retrieval interference in which expectation effects play out in early, lexical processing while dependency integration processes occur later without interacting with expectation.
Advisors/Committee Members: Lewis, Richard L. (committee member), Boland, Julie (committee member), Meyer, David E. (committee member), Vasishth, Shravan (committee member).
Subjects/Keywords: Working Memory; Sentence Processing; Expectation; Similarity-based Interference; Locality; Psychology; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Bartek, B. D. (2011). Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/86317
Chicago Manual of Style (16th Edition):
Bartek, Brian D. “Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context.” 2011. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/86317.
MLA Handbook (7th Edition):
Bartek, Brian D. “Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context.” 2011. Web. 04 Mar 2021.
Vancouver:
Bartek BD. Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context. [Internet] [Doctoral dissertation]. University of Michigan; 2011. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/86317.
Council of Science Editors:
Bartek BD. Short-term Memory Retrievals and Expectation in On-line Sentence Comprehension: The Effects of Recent Linguistic Context. [Doctoral Dissertation]. University of Michigan; 2011. Available from: http://hdl.handle.net/2027.42/86317
12.
Gorski, Nicholas A.
Learning To Use Memory.
Degree: PhD, Computer Science & Engineering, 2012, University of Michigan
URL: http://hdl.handle.net/2027.42/91491
► This thesis is a comprehensive empirical exploration of using reinforcement learning to learn to use simple forms of working memory. Learning to use memory involves…
(more)
▼ This thesis is a comprehensive empirical exploration of using reinforcement learning to learn to use simple forms of working memory. Learning to use memory involves learning how to behave in the environment while simultaneously learning when to select internal actions that control how knowledge persists in memory and learning how to use that information stored in memory to make decisions. We focus on two different models of memory: bit memory and gated memory. Bit memory is inspired by prior reinforcement learning literature and stores abstract values, which an agent can learn to associate with task history. Gated memory is inspired by human working memory and stores perceptually grounded symbols. Our goal is to determine computational bounds on the tractability of learning to use these memories. We conduct a comprehensive empirical exploration of the dynamics of learning to use memory models by modifying a simple partially observable task, TMaze, along specific dimensions: length of temporal delay, number of dependent decisions, number of distinct symbols, quantity of concurrent knowledge, and availability of second-order knowledge. We find that learning to use gated memory is significantly more tractable than learning to use bit memory because it stores perceptually grounded symbols in memory. We further find that learning performance scales more favorably along temporal delay, distinct symbols, and concurrent knowledge when learning to use gated memory than along other dimensions. We also identify situations in which agents fail to learn to use gated memory optimally which involve repeated identical observations which result in no unambiguous trajectories through the underlying task and memory state space.
Advisors/Committee Members: Laird, John E. (committee member), Baveja, Satinder Singh (committee member), Lewis, Richard L. (committee member), Polk, Thad A. (committee member).
Subjects/Keywords: Reinforcement Learning; Memory; Artificial Intelligence; Sequential Decision Making; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Gorski, N. A. (2012). Learning To Use Memory. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/91491
Chicago Manual of Style (16th Edition):
Gorski, Nicholas A. “Learning To Use Memory.” 2012. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/91491.
MLA Handbook (7th Edition):
Gorski, Nicholas A. “Learning To Use Memory.” 2012. Web. 04 Mar 2021.
Vancouver:
Gorski NA. Learning To Use Memory. [Internet] [Doctoral dissertation]. University of Michigan; 2012. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/91491.
Council of Science Editors:
Gorski NA. Learning To Use Memory. [Doctoral Dissertation]. University of Michigan; 2012. Available from: http://hdl.handle.net/2027.42/91491
13.
Tutunjian, Damon A.
Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism.
Degree: PhD, Linguistics, 2010, University of Michigan
URL: http://hdl.handle.net/2027.42/78841
► This dissertation examines the influence of lexical-semantic representations, conceptual similarity, and contextual fit on the processing of coordinated verb phrases. The study integrates information gleaned…
(more)
▼ This dissertation examines the influence of lexical-semantic representations, conceptual similarity, and contextual fit on the processing of coordinated verb phrases. The study integrates information gleaned from current linguistic theory with current psycholinguistic approaches to examining the processing of coordinated verb phrases.
It has been claimed that in coordinated phrases, one conjunct may influence the processing of a second conjunct if they are sufficiently similar. For example, The likelihood of adopting an intransitive analysis for the optionally transitive verb of a subordinated clause in sentences like "Although the pirate ship sank the nearby British vessel did not send out lifeboats" may be increased if the ambiguous verb ("sank") is coordinated with a preceding, intransitively biased verb ("halted and sank"). Similarly, processing of the second conjunct may be facilitated when coordinated with a similar first conjunct. Such effects, and others in this vein have often been designated “parallelism effects.”
However, notions of similarity underlying such effects have long been ill-defined. Many existing studies rely on relatively shallow features like syntactic category information or argument structure generalizations, such as transitive or intransitive, as a basis for structural comparison. But it may be that deeper levels of lexical-semantic representation and more varied, semantic or conceptual sources of information are also relevant to establishing similarity between conjuncts. In addition, little has been done to integrate parallelism effects to theories of the processing architecture underlying such effects, particularly for studies involving syntactic ambiguity resolution.
Using two word-by-word reading and three eyetracking while reading experiments, I investigate what contribution detailed lexical-semantic representations, as well as conceptual and contextual information make towards establishing parallel coordination in the online processing of coordinated verb phrases. The five studies demonstrate that parallelism effects are indeed sensitive to deeper representational information, conceptual similarity, and contextual fit. Furthermore, by controlling for deeper representational information, it is demonstrated that expected facilitatory patterns arising from coordination of similar conjuncts may be disrupted. Implications for the architecture of the processing system are discussed, and it is argued that constraint-based/competition models of processing best accommodate the pattern of results.
Advisors/Committee Members: Boland, Julie (committee member), Epstein, Samuel D. (committee member), Keshet, Ezra Russell (committee member), Lewis, Richard L. (committee member).
Subjects/Keywords: Sentence Processing; Eyetracking; Coordination; Argument Structure; Lexical Semantics; Linguistics; Psychology; Humanities; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Tutunjian, D. A. (2010). Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/78841
Chicago Manual of Style (16th Edition):
Tutunjian, Damon A. “Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism.” 2010. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/78841.
MLA Handbook (7th Edition):
Tutunjian, Damon A. “Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism.” 2010. Web. 04 Mar 2021.
Vancouver:
Tutunjian DA. Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism. [Internet] [Doctoral dissertation]. University of Michigan; 2010. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/78841.
Council of Science Editors:
Tutunjian DA. Processing Coordinated Verb Phrases: The Relevance of Lexical-Semantic, Conceptual, and Contextual Information towards Establishing Verbal Parallelism. [Doctoral Dissertation]. University of Michigan; 2010. Available from: http://hdl.handle.net/2027.42/78841
14.
Guo, Xiaoxiao.
Deep Learning and Reward Design for Reinforcement Learning.
Degree: PhD, Computer Science & Engineering, 2017, University of Michigan
URL: http://hdl.handle.net/2027.42/136931
► One of the fundamental problems in Artificial Intelligence is sequential decision making in a flexible environment. Reinforcement Learning (RL) gives a set of tools for…
(more)
▼ One of the fundamental problems in Artificial Intelligence is sequential decision making in a flexible environment. Reinforcement Learning (RL) gives a set of tools for solving sequential decision problems. Although the theory of RL addresses a general class of learning problems with a constructive mathematical formulation, the challenges posed by the interaction of rich perception and delayed rewards in many domains remain a significant barrier to the widespread applicability of RL methods.
The rich perception problem itself has two components: 1) the sensors at any time step do not capture all the information in the history of observations, leading to partial observability, and 2) the sensors provide very high-dimensional observations, such as images and natural languages, that introduce computational and sample-complexity challenges for the representation and generalization problems in policy selection. The delayed reward problem—that the effect of actions in terms of future rewards is delayed in time—makes it hard to determine how to credit action sequences for reward outcomes.
This dissertation offers a set of contributions that adapt the hierarchical representation learning power of deep learning to address rich perception in vision and text domains, and develop new reward design algorithms to address delayed rewards. The first contribution is a new learning method for deep neural networks in vision-based real-time control. The learning method distills slow policies of the Monte Carlo Tree Search (MCTS) into fast convolutional neural networks, which outperforms the conventional Deep Q-Network. The second contribution is a new end-to-end reward design algorithm to mitigate the delayed rewards for the state-of-the-art MCTS method. The reward design algorithm converts visual perceptions into reward bonuses via deep neural networks, and optimizes the network weights to improve the performance of MCTS end-to-end via policy gradient. The third contribution is to extend existing policy gradient reward design method from single task to multiple tasks. Reward bonuses learned from old tasks are transferred to new tasks to facilitate learning. The final contribution is an application of deep reinforcement learning to another type of rich perception, ambiguous texts. A synthetic data set is proposed to evaluate the querying, reasoning and question-answering abilities of RL agents, and a deep memory network architecture is applied to solve these challenging problems to substantial degrees.
Advisors/Committee Members: Baveja, Satinder Singh (committee member), Lewis, Richard L (committee member), Mei, Qiaozhu (committee member), Lee, Honglak (committee member).
Subjects/Keywords: Reinforcement Learning; Deep Learning; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Guo, X. (2017). Deep Learning and Reward Design for Reinforcement Learning. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/136931
Chicago Manual of Style (16th Edition):
Guo, Xiaoxiao. “Deep Learning and Reward Design for Reinforcement Learning.” 2017. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/136931.
MLA Handbook (7th Edition):
Guo, Xiaoxiao. “Deep Learning and Reward Design for Reinforcement Learning.” 2017. Web. 04 Mar 2021.
Vancouver:
Guo X. Deep Learning and Reward Design for Reinforcement Learning. [Internet] [Doctoral dissertation]. University of Michigan; 2017. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/136931.
Council of Science Editors:
Guo X. Deep Learning and Reward Design for Reinforcement Learning. [Doctoral Dissertation]. University of Michigan; 2017. Available from: http://hdl.handle.net/2027.42/136931
15.
Hsieh, Yu-Fen.
Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution.
Degree: PhD, Linguistics, 2010, University of Michigan
URL: http://hdl.handle.net/2027.42/77872
► Four reading-time studies in the dissertation investigated the online representation of a syntactic ambiguity and the nature of the time course of the interaction between…
(more)
▼ Four reading-time studies in the dissertation investigated the online representation of a syntactic ambiguity and the nature of the time course of the interaction between syntactic and non-syntactic constraints. The target syntactic ambiguity was the construction of Verb NP1 de NP2 in Chinese, which is ambiguous between a relative clause (RC) and a complement clause (CC) analysis.
Using an eye-tracking paradigm, Experiments 1 and 2 explored whether the parser can maintain multiple alternative structures of an ambiguity and how semantic plausibility influences the early stage of syntactic processing. The results demonstrated that the degree of processing difficulty at the disambiguation varied as a function of the relative support for the RC and the CC alternatives from the syntactic and the semantic constraints. The findings can be best accounted for by a limited, ranked parallel parsing model, such as the surprisal theory (Hale, 2001), which maintains that processing difficulty is incurred by resource reallocation during disambiguation.
Experiments 3 and 4 utilized syntactic priming to investigate how recent prior experience with a particular structure can influence syntactic ambiguity resolution in comprehension. Experiment 3 showed that lexically independent priming in comprehension facilitated the accessibility of the repeated RC structure, increasing the difficulty of structural revision to the unprimed CC alternative. Experiment 4 found that prior experience with the English RC structure affected the processing of the corresponding structure in Chinese, even though the RC structures differ in word order in the two languages. The observed syntactic priming in comprehension between Chinese and English RC structures suggested that the two languages have a shared syntactic representation that does not specify word order.
Overall, the dissertation contributes to the understanding of structural representation and information integration during syntactic ambiguity resolution. The findings provided evidence for an interactive and limited parallel approach to sentence processing. Moreover, lexically independent comprehension priming suggested that prior experience with a particular syntactic configuration can function as a constraint at the structural level. Thus, a traditional constraint-based lexicalist theory (e.g. MacDonald, Pearlmutter, & Seidenberg, 1994) must incorporate non-lexical representations in order to make use of statistical regularities beyond the lexical level.
Advisors/Committee Members: Boland, Julie (committee member), Duanmu, San (committee member), Lewis, Richard L. (committee member), Pires, Acrisio (committee member).
Subjects/Keywords: Psycholinguistics; Sentence Processing; Syntactic Ambiguity Resolution; Eye-tracking; Syntactic Priming; Bilingual Sentence Processing; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Hsieh, Y. (2010). Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/77872
Chicago Manual of Style (16th Edition):
Hsieh, Yu-Fen. “Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution.” 2010. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/77872.
MLA Handbook (7th Edition):
Hsieh, Yu-Fen. “Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution.” 2010. Web. 04 Mar 2021.
Vancouver:
Hsieh Y. Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution. [Internet] [Doctoral dissertation]. University of Michigan; 2010. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/77872.
Council of Science Editors:
Hsieh Y. Sentence Processing in Chinese and Chinese-English Bilinguals: Syntax-Semantics Interaction During Syntactic Ambiguity Resolution. [Doctoral Dissertation]. University of Michigan; 2010. Available from: http://hdl.handle.net/2027.42/77872
16.
Chou, Chao-Ting.
Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese.
Degree: PhD, Linguistics, 2013, University of Michigan
URL: http://hdl.handle.net/2027.42/99859
► This dissertation reveals an apparent paradox presented by Chinese which confronts current theories of Universal Grammar, parametric variation, and typology of human languages. Specifically, Chinese…
(more)
▼ This dissertation reveals an apparent paradox presented by Chinese which confronts current theories of Universal Grammar, parametric variation, and typology of human languages. Specifically, Chinese presents a problem for the feature inheritance hypothesis (Richards 2007; Chomsky 2007, 2008; Miyagawa 2010) because Chinese seems to manifest neither phi-features/Agree on T nor the freedom to raise any topic phrase to spec-TP (a hallmark feature of discourse-configurational languages like Finnish). That is, the UG-typology advanced under the feature inheritance hypotheses seems to exclude Chinese.
In view of this apparent paradox, I investigate the distribution and motivation of A-movement in Chinese in chapter 2 and argue that Chinese is in fact compatible with Miyagawa’s feature inheritance approach. I contend that Chinese displays A-movement motivated by two distinct forces: the Case feature (see Epstein and Seely 2006, Bošković 2002) and the Topic feature (as in Miyagawa 2010).
Chapter 3 investigates the question of whether Chinese lacks phi-features/Agree altogether by examining the blocking effects observed in the long-distance construal of the reflexive ziji (see Huang and Liu 2001) and the formation of Chinese wh-the-hell questions (see Huang and Ochi 2004 and Chou 2012). I contend that these two types of blocking effects receive a unified analysis if we assume phi-Agree exists at the CP level in Chinese, and involves [Speaker] and [Participant] features.
Chapter 4 examines the derivation of locative inversion in English and Chinese, particularly focusing on why T-to-C inversion is not allowed in English locative inversion, whereas the counterpart operation is allowed in Chinese. I argue that (i) locative inversion in both English and Chinese is topic A-movement, and (ii) the possible presence of φ-features on T and the categorial status of the locative phrase jointly determine whether a language can implement T-to-C inversion in locative inversion. Chapter 5 discusses two foundational theoretical implications, including (i) how to express the A/A'-distinction in languages without φ-features on T, and (ii) the postulation of featurally crash-proof grammar in which uninterpretable features present at the CI interface do not cause crash (see Frampton and Guttman 2002; Carstens 2011; Epstein, Kitahara and Seely 2010; Putnam 2010).
Advisors/Committee Members: Pires, Acrisio (committee member), Epstein, Samuel D. (committee member), Lewis, Richard L. (committee member), Seely, T. Daniel (committee member), Keshet, Ezra Russell (committee member).
Subjects/Keywords: Universal Grammar; Parameters; Minimalist Syntax; Chinese Syntax; A-movement; Phi-features/Agreement; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Chou, C. (2013). Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/99859
Chicago Manual of Style (16th Edition):
Chou, Chao-Ting. “Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese.” 2013. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/99859.
MLA Handbook (7th Edition):
Chou, Chao-Ting. “Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese.” 2013. Web. 04 Mar 2021.
Vancouver:
Chou C. Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese. [Internet] [Doctoral dissertation]. University of Michigan; 2013. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/99859.
Council of Science Editors:
Chou C. Phi-Agree, A-movement, and Complementizer-Tense Relations in Chinese. [Doctoral Dissertation]. University of Michigan; 2013. Available from: http://hdl.handle.net/2027.42/99859
17.
Lent, Jeremy A. B.
Knowing by Example: A Social-Cognitive Approach to Epistemology.
Degree: PhD, Philosophy, 2016, University of Michigan
URL: http://hdl.handle.net/2027.42/133254
► In this dissertation, I explore the “problem of knowledge” in epistemology from a new angle. Rather than proposing conditions for when it is true or…
(more)
▼ In this dissertation, I explore the “problem of knowledge” in epistemology from a new angle. Rather than proposing conditions for when it is true or false that someone’s belief constitutes knowledge, I take as my challenge the problem of explaining why we use the words “know” and “knowledge” in the ways that we do. Ever since Edmund Gettier’s 1963 paper “Is Justified True Belief Knowledge?”, epistemologists have uncovered many puzzling aspects of how we use these two words. In Chapter 2, I hypothesize that our human ancestors developed a word to urge others to be certain (or not) of particular facts. Because this word was understood to invoke objective reasons, the message conveyed was not only “I advise us to be certain that such-and-such” but also (at least implicitly) “I would advise anyone in a position similar to ours to do likewise”. I suggest that our word “know” is the descendant of this ancestral word, and continues to carry a “prospective” advisory message. I show that this hypothesis accounts for several features of our word “know”, such as that we apply it only when truth, certainty, and justification are present. In Chapters 3 and 4, I use the hypothesis to explain why philosophers have the verdicts they do on many puzzling cases of “true belief without knowledge” that have been discussed in the epistemology literature. In Chapter 5, I use my hypothesis to explain the results of several recent experiments conducted to test non-philosophers’ judgments about these same cases. In Chapter 6, I conclude that my hypothesis is the best available explanation for all of the puzzling features of our modern-day uses of “know”. We should thereby provisionally accept that my hypothesis (or something like it) gives the truth about the development and functioning of our word “know”. Although my hypothesis belongs to the domains of evolutionary biology and social-cognitive psychology, I argue that it also counts as a piece of philosophy: It helps us to understand puzzling features of our lives, not for any instrumental reason but simply for the sake of understanding.
Advisors/Committee Members: Anderson, Elizabeth S (committee member), Joyce, James M (committee member), Lewis, Richard L (committee member), Railton, Peter A (committee member), Aarnio, Maria Lasonen (committee member).
Subjects/Keywords: epistemology; analysis of knowledge; Gettier problem; Philosophy; Psychology; Humanities; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Lent, J. A. B. (2016). Knowing by Example: A Social-Cognitive Approach to Epistemology. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/133254
Chicago Manual of Style (16th Edition):
Lent, Jeremy A B. “Knowing by Example: A Social-Cognitive Approach to Epistemology.” 2016. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/133254.
MLA Handbook (7th Edition):
Lent, Jeremy A B. “Knowing by Example: A Social-Cognitive Approach to Epistemology.” 2016. Web. 04 Mar 2021.
Vancouver:
Lent JAB. Knowing by Example: A Social-Cognitive Approach to Epistemology. [Internet] [Doctoral dissertation]. University of Michigan; 2016. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/133254.
Council of Science Editors:
Lent JAB. Knowing by Example: A Social-Cognitive Approach to Epistemology. [Doctoral Dissertation]. University of Michigan; 2016. Available from: http://hdl.handle.net/2027.42/133254
18.
Derbinsky, Nathaniel Leonard.
Effective and Efficient Memory for Generally Intelligent Agents.
Degree: PhD, Computer Science and Engineering, 2012, University of Michigan
URL: http://hdl.handle.net/2027.42/93855
► Intelligent systems with access to large stores of experience, or memory, can draw upon and reason about this knowledge in a variety of situations, such…
(more)
▼ Intelligent systems with access to large stores of experience, or memory, can draw upon and reason about this knowledge in a variety of situations, such as to improve the efficacy of their learning, decision-making, and actions in the world. However, little research has examined the computational challenges that arise when real-time agents require access to large stores of knowledge over long periods of time.
This dissertation explores the computational trade-offs involved in enhancing intelligent agents with effective and efficient memory. We exploit general properties of environments, tasks, and agent cues in order to develop scalable algorithms for episodic learning (autobiographical memory); semantic learning (context-independent store of facts and relations); and competence-preserving retention of learned knowledge (policies to forget memories while maintaining task performance). We evaluate these algorithms in Soar, a general cognitive architecture, for hours-to-days of real-time execution and demonstrate that agents with effective and efficient memory benefit along numerous dimensions when tasked within a variety of problem domains, including linguistics, planning, games, and mobile robotics.
Advisors/Committee Members: Laird, John E. (committee member), Lewis, Richard L. (committee member), Cafarella, Michael John (committee member), Kuipers, Benjamin (committee member), Van Lent, Michael C. (committee member).
Subjects/Keywords: Artificial Intelligence; Episodic Memory; Semantic Memory; Forgetting; Soar; Cognitive Architecture; Computer Science; Engineering
…University of Michigan is located in Ann Arbor ), whereas episodic
memories allow an…
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Derbinsky, N. L. (2012). Effective and Efficient Memory for Generally Intelligent Agents. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/93855
Chicago Manual of Style (16th Edition):
Derbinsky, Nathaniel Leonard. “Effective and Efficient Memory for Generally Intelligent Agents.” 2012. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/93855.
MLA Handbook (7th Edition):
Derbinsky, Nathaniel Leonard. “Effective and Efficient Memory for Generally Intelligent Agents.” 2012. Web. 04 Mar 2021.
Vancouver:
Derbinsky NL. Effective and Efficient Memory for Generally Intelligent Agents. [Internet] [Doctoral dissertation]. University of Michigan; 2012. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/93855.
Council of Science Editors:
Derbinsky NL. Effective and Efficient Memory for Generally Intelligent Agents. [Doctoral Dissertation]. University of Michigan; 2012. Available from: http://hdl.handle.net/2027.42/93855
19.
Aronowitz, Sarah.
Rational Structures in Learning and Memory.
Degree: PhD, Philosophy, 2018, University of Michigan
URL: http://hdl.handle.net/2027.42/145896
► My dissertation aims to disrupt an increasingly ubiquitous view of epistemology which claim that we can study rationality by considering a single belief at a…
(more)
▼ My dissertation aims to disrupt an increasingly ubiquitous view of epistemology which claim that we can study rationality by considering a single belief at a single time. I target three areas where diachronic (i.e. temporal) factors make a difference in the three sections:
1. memory, a system of tremendous importance in our cognitive lives yet which is often reduced to a one-sided question of whether to trust what one’s memory says,
2. learning, where I argue that we should sometimes believe in a way that’s not warranted or reasonable in light of our current evidence, but which puts us in a better position to acquire evidence in the future, and
3. the connection between memory and learning, as exemplified in the case of remembering anomalous events.
This project is important because our whole lives are organized around getting things right at the right time. When we try to act morally, we might try to have a life that is built around moral principles, or to become wiser and kinder over time, as opposed to amassing a collection of acts that all have independent moral value. I think the same thing is true of our endeavors to acquire knowledge the process of inquiry is not made up of individual, independent good inferences that happen to follow one another, but is instead about a trajectory where we learn over time, and take the right steps now to get things right in the future, and overall. So I think that to understand this more complete sense of inquiry, philosophy needs to make a place for memory, the system that sustains and directs inquiry in the background, over long periods of time even as the sciences are learning more and more about how natural memory systems work, philosophers have boxed it out of relevance.My methodology is to study natural and artificial learning and memory systems as a process of discovery, a way of using real-world cases as inspiration and guide to the
normative landscape. Conversely, I hope that figuring out new normative possibilities can shed light on empirical facts - though this is not the main focus of my dissertation.
Advisors/Committee Members: Railton, Peter A (committee member), Lewis, Richard L (committee member), Joyce, James M (committee member), Sripada, Sekhar Chandra (committee member), Williamson, Timothy (committee member).
Subjects/Keywords: epistemology; memory; diachronic rationality; Philosophy; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Aronowitz, S. (2018). Rational Structures in Learning and Memory. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/145896
Chicago Manual of Style (16th Edition):
Aronowitz, Sarah. “Rational Structures in Learning and Memory.” 2018. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/145896.
MLA Handbook (7th Edition):
Aronowitz, Sarah. “Rational Structures in Learning and Memory.” 2018. Web. 04 Mar 2021.
Vancouver:
Aronowitz S. Rational Structures in Learning and Memory. [Internet] [Doctoral dissertation]. University of Michigan; 2018. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/145896.
Council of Science Editors:
Aronowitz S. Rational Structures in Learning and Memory. [Doctoral Dissertation]. University of Michigan; 2018. Available from: http://hdl.handle.net/2027.42/145896
20.
Wang, Yongjia.
Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments.
Degree: PhD, Computer Science & Engineering, 2011, University of Michigan
URL: http://hdl.handle.net/2027.42/84437
► Creating autonomous long-lived agent that can robustly function in a complex object- based environment has been a persistent goal in the field of artificial intelligence.…
(more)
▼ Creating autonomous long-lived agent that can robustly function in a complex object- based environment has been a persistent goal in the field of artificial intelligence. Learning the appropriate functional categories of objects is one of the keys to achieve this goal, and is the theme of this thesis.
We formulate the research problem as finding efficient value function approximation algorithms, where the input to the function is an object-based state representation, and output of the function is the utility value of that input state. The challenges arise from the requirements of efficient learning, and incremental learning of complex nonlinear value functions, whose input consists of diverse objects in high dimensional feature space. Our solutions are based on three key principles. The first is that the value function representation can usefully exploit the compositional structure of object-based environments, where the state representations consist of independent objects with their own perceptual features and functional properties. The second is that hierarchical symbolic category representations, inspired by human cognitive models, can help achieve efficient learning. The third is that the object categorization criteria must be consistent and coherent with the target utility value function. We provide two implementations based on these key principles, with evaluations both based on functionality and on cognitive plausibility.
Traditionally, category learning and value function approximation are studied as separate problems. The thesis presents a unique synthesis of the two. On one hand, it provides efficient value function approximation algorithms that can take advantage of compact representational basis adaptively generated by hierarchical category learning. On the other hand, it provides a utility based category learning model that offers new computational insights to human category learning behaviors.
Advisors/Committee Members: Laird, John E. (committee member), Baveja, Satinder Singh (committee member), Kuipers, Benjamin (committee member), Lewis, Richard L. (committee member), Polk, Thad A. (committee member).
Subjects/Keywords: Category Learning; Category Formation; Hierarchical Category; Functional Category; Cognitive Architecture; Intelligent Agent; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Wang, Y. (2011). Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/84437
Chicago Manual of Style (16th Edition):
Wang, Yongjia. “Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments.” 2011. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/84437.
MLA Handbook (7th Edition):
Wang, Yongjia. “Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments.” 2011. Web. 04 Mar 2021.
Vancouver:
Wang Y. Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments. [Internet] [Doctoral dissertation]. University of Michigan; 2011. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/84437.
Council of Science Editors:
Wang Y. Hierarchical Functional Category Learning for Efficient Value Function Approximation in Object-Based Environments. [Doctoral Dissertation]. University of Michigan; 2011. Available from: http://hdl.handle.net/2027.42/84437
21.
Nediger, Will.
Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms.
Degree: PhD, Linguistics, 2017, University of Michigan
URL: http://hdl.handle.net/2027.42/138471
► Idioms have traditionally posed difficulties for different syntactic frameworks, because they behave in some senses like lexical items but in other senses like syntactically complex…
(more)
▼ Idioms have traditionally posed difficulties for different syntactic frameworks, because they behave in some senses like lexical items but in other senses like syntactically complex phrases. In particular, despite showing evidence of having internal syntactic structure, they have apparently limited syntactic flexibility relative to non-idiomatic phrases. This dissertation proposes a Minimalist architecture which makes a sharp distinction between the lexicon and the syntax, but nonetheless accounts for the hybrid properties of idioms. I argue that idioms, like non-idiomatic structures, are built by iterative application of Merge, preserving the Minimalist notion that there is a single basic structure-building operation, Merge, in natural language. However, idioms are also stored wholesale in the lexicon in the form of syntactic structures with associated phonological and semantic representations. These lexically stored idioms do not serve as input to structure building through Merge. Rather, if the syntactic derivation builds a structure which matches a lexically stored idiom, then that structure may optionally be interpreted via the lexically stored idiom meaning.
Given my proposal that all idioms are built by means of Merge, I analyze extensive evidence for syntactic flexibility across different types of idioms, and argue that the apparent limitations on the syntactic flexibility of idioms can be explained without positing any idiom-specific restrictions. Rather, I explain how the conceptual-intentional interface imposes independent semantic restrictions that constrain the syntactic derivation of particular idioms, accounting for distinctions that include the much-discussed contrast between decomposable idioms (whose meaning is distributed among their parts, e.g. spill the beans, in which spill can be paraphrased as ‘divulge’ and beans can be paraphrased as ‘secret’) and non-decomposable idioms (whose meaning is not distributed among their parts, e.g. kick the bucket, in which no independent meaning can be identified for kick or bucket). The semantic representations I propose for non-decomposable idioms are associated with their entire lexically stored structure, unlike those for decomposable idioms. This distinction interacts with independent semantic constraints to explain the apparently limited syntactic flexibility of non-decomposable idioms relative to decomposable idioms. This approach extends to idioms a unified structure-building procedure for natural language, while explaining the linguistic properties of idioms in a principled way, consistent with Minimalist assumptions.
Advisors/Committee Members: Pires, Acrisio M (committee member), Lewis, Richard L (committee member), Baptista, Marlyse (committee member), Epstein, Samuel D (committee member), Keshet, Ezra Russell (committee member).
Subjects/Keywords: idioms; syntax-semantics interface; Minimalist Program; Merge; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Nediger, W. (2017). Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/138471
Chicago Manual of Style (16th Edition):
Nediger, Will. “Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms.” 2017. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/138471.
MLA Handbook (7th Edition):
Nediger, Will. “Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms.” 2017. Web. 04 Mar 2021.
Vancouver:
Nediger W. Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms. [Internet] [Doctoral dissertation]. University of Michigan; 2017. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/138471.
Council of Science Editors:
Nediger W. Unifying Structure-Building in Human Language: The Minimalist Syntax of Idioms. [Doctoral Dissertation]. University of Michigan; 2017. Available from: http://hdl.handle.net/2027.42/138471
22.
Mohan, Shiwali.
From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction.
Degree: PhD, Computer Science and Engineering, 2015, University of Michigan
URL: http://hdl.handle.net/2027.42/111573
► Intelligent collaborative agents are becoming common in the human society. From virtual assistants such as Siri and Google Now to assistive robots, they contribute to…
(more)
▼ Intelligent collaborative agents are becoming common in the human society. From virtual assistants such as Siri and Google Now to assistive robots, they contribute to human activities in a variety of ways. As they become more pervasive, the challenge of customizing them to a variety of environments and tasks becomes critical. It is infeasible for engineers to program them for each individual use. Our research aims at building interactive robots and agents that adapt to new environments autonomously by interacting with human users using natural modalities.
This dissertation studies the problem of learning novel tasks from human-agent dialog. We propose a novel approach for interactive task learning, situated interactive instruction (SII), and investigate approaches to three computational challenges that arise in designing SII agents: situated comprehension, mixed-initiative interaction, and interactive task learning. We propose a novel mixed-modality grounded representation for task verbs which encompasses their lexical, semantic, and
task-oriented aspects. This representation is useful in situated comprehension and can be learned through human-agent interactions. We introduce the Indexical Model of comprehension that can exploit
extra-linguistic contexts for resolving semantic ambiguities in situated comprehension of task commands. The Indexical model is integrated with a mixed-initiative interaction model that facilitates
a flexible task-oriented human-agent dialog. This dialog serves as the basis of interactive task learning. We propose an interactive variation of explanation-based learning that can acquire the proposed
representation. We demonstrate that our learning paradigm is efficient, can transfer knowledge between structurally similar tasks, integrates agent-driven exploration with instructional learning, and can acquire several tasks. The methods proposed in this thesis are integrated in Rosie - a generally instructable agent developed in the Soar cognitive architecture and embodied on a table-top robot.
Advisors/Committee Members: Laird, John E. (committee member), Lewis, Richard L. (committee member), Thomaz, Andrea Lockerd (committee member), Olson, Edwin (committee member), Durfee, Edmund H. (committee member).
Subjects/Keywords: interactive task learning; situated interactive instruction; human-robot interaction; robot learning; grounded language comprehension; interactive agents; Chemical Engineering; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Mohan, S. (2015). From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/111573
Chicago Manual of Style (16th Edition):
Mohan, Shiwali. “From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction.” 2015. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/111573.
MLA Handbook (7th Edition):
Mohan, Shiwali. “From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction.” 2015. Web. 04 Mar 2021.
Vancouver:
Mohan S. From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction. [Internet] [Doctoral dissertation]. University of Michigan; 2015. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/111573.
Council of Science Editors:
Mohan S. From Verbs to Tasks: An Integrated Account of Learning Tasks from Situated Interactive Instruction. [Doctoral Dissertation]. University of Michigan; 2015. Available from: http://hdl.handle.net/2027.42/111573
23.
Obata, Miki.
Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer.
Degree: PhD, Linguistics, 2010, University of Michigan
URL: http://hdl.handle.net/2027.42/77877
► The goal of the dissertation is to determine aspects of the structure of the human language faculty, a cognitive system, specifically focusing on human syntactic…
(more)
▼ The goal of the dissertation is to determine aspects of the structure of the human language faculty, a cognitive system, specifically focusing on human syntactic systems, (unique in the animal kingdom) which enable us to creatively produce an unlimited number of grammatical sentences (like the one you just read, probably never before written or thought). The thesis especially addresses mechanisms of the syntactic movement operation, namely Internal Merge, a central concern in the field. I analyze the data from a cross-linguistic perspective and investigate how the language faculty can compute the validity of movement under Chomsky's (2007, 2008) analyses.
The main focus is on the following two points: [1] (re-)formulating fundamental mechanisms of the computational system especially concerning the operation Merge, which assembles two elements (i.e. each a bundle of features) creating a new syntactic object, a set, and [2] addressing certain central issues of how syntactic representations (constructed by Merge) are (cyclically) transferred from the narrow syntax to the interface components, in which the semantic and phonological computations are carried out.
The thesis is mainly devoted to [1] and [2] within the framework of the phase-based derivational approach suggested in Chomsky (2000, 2001, 2004, 2007, 2008), where the syntactic computation is 'chunked' i.e. limited to small units called phases, by assumption, leading to the reduction of computational complexity. With respect to [1], I closely examine the case that multiple heads are involved in triggering Internal Merge (i.e. movement) of a single element under the feature-inheritance system, introduced in Chomsky (2007, 2008) and propose Feature-Splitting Internal Merge, which enables us to explain improper movement phenomena within local computation, in contrast to May (1979). Regarding [2], the mechanism of Transfer is discussed, especially focusing on the timing of its application and also seeking to explain which portion (category) within the constructed narrow syntax representation it applies to. In addition, the thesis will discuss how the narrow syntax is constrained by the interface conditions.
Advisors/Committee Members: Chomsky, Noam (committee member), Epstein, Samuel D. (committee member), Kitahara, Hisatsugu (committee member), Lewis, Richard L. (committee member), Pires, Acrisio (committee member), Seely, T. Daniel (committee member).
Subjects/Keywords: Generative Grammar; Minimalist Program; Syntactic Movement; Phase-based Derivational Approach; Cognitive Science; Linguistics; Humanities
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Obata, M. (2010). Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/77877
Chicago Manual of Style (16th Edition):
Obata, Miki. “Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer.” 2010. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/77877.
MLA Handbook (7th Edition):
Obata, Miki. “Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer.” 2010. Web. 04 Mar 2021.
Vancouver:
Obata M. Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer. [Internet] [Doctoral dissertation]. University of Michigan; 2010. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/77877.
Council of Science Editors:
Obata M. Root, Successive-Cyclic and Feature-Splitting Internal Merge: Implications for Feature-Inheritance and Transfer. [Doctoral Dissertation]. University of Michigan; 2010. Available from: http://hdl.handle.net/2027.42/77877
24.
Xu, Joseph Zhen Ying.
Learning Integrated Relational and Continuous Action Models for Continuous Domains.
Degree: PhD, Computer Science & Engineering, 2013, University of Michigan
URL: http://hdl.handle.net/2027.42/102383
► Long-living autonomous agents must be able to learn to perform competently in novel environments. One important aspect of competence is the ability to plan, which…
(more)
▼ Long-living autonomous agents must be able to learn to perform competently in
novel environments. One important aspect of competence is the ability to plan,
which entails the ability to learn models of the agent's own actions and their
effects on the environment. This thesis describes an approach to learn action
models of environments with continuous-valued spatial states and realistic
physics consisting of multiple interacting rigid objects. In such environments,
we hypothesize that objects exhibit multiple qualitatively distinct behaviors
based on their relationships to each other and how they interact. We call
these qualitatively distinct behaviors modes. Our approach models individual
modes with linear functions. We extend the standard propositional function
representation with learned knowledge about the roles of objects in
determining the outcomes of functions. Roles are learned as first-order
relations using the FOIL algorithm. This allows the functions modeling
individual modes to be "instantiated" with different sets of objects, similar
to relational rules such as STRIPS operators. We also use FOIL to learn
preconditions for each mode consisting of clauses that test spatial
relationships between objects. These relational preconditions naturally
capture the interaction dynamics of spatial domains and allow faster learning
and generalization of the model. The combination of continuous linear
functions, relational roles, and relational mode preconditions effectively
capture both continuous and relational regularities prominent in spatial
domains. This results in faster and more general action modeling in these
domains. We evaluate the algorithm on two domains, one involving pushing
stacks of boxes against frictional resistance, and one in which a ball
interacts with obstacles in a physics simulator. We show that our algorithm
learns more accurate models than locally weighted regression in the physics
simulator domain. We also show that relational mode preconditions learned with
FOIL are more accurate than continuous classifiers learned with support vector
machines and k-nearest-neighbor.
Advisors/Committee Members: Laird, John E. (committee member), Lewis, Richard L. (committee member), Kuipers, Benjamin (committee member), Lee, Honglak (committee member), Baveja, Satinder Singh (committee member).
Subjects/Keywords: Action Modeling; Learning; Combined Symbolic/Continuous Representation; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Xu, J. Z. Y. (2013). Learning Integrated Relational and Continuous Action Models for Continuous Domains. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/102383
Chicago Manual of Style (16th Edition):
Xu, Joseph Zhen Ying. “Learning Integrated Relational and Continuous Action Models for Continuous Domains.” 2013. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/102383.
MLA Handbook (7th Edition):
Xu, Joseph Zhen Ying. “Learning Integrated Relational and Continuous Action Models for Continuous Domains.” 2013. Web. 04 Mar 2021.
Vancouver:
Xu JZY. Learning Integrated Relational and Continuous Action Models for Continuous Domains. [Internet] [Doctoral dissertation]. University of Michigan; 2013. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/102383.
Council of Science Editors:
Xu JZY. Learning Integrated Relational and Continuous Action Models for Continuous Domains. [Doctoral Dissertation]. University of Michigan; 2013. Available from: http://hdl.handle.net/2027.42/102383
25.
Shvartsman, Michael.
Adaptive Eye Movement Control in a Simple Linguistic Task.
Degree: PhD, Psychology, 2014, University of Michigan
URL: http://hdl.handle.net/2027.42/110380
► This dissertation pursues a computationally rational analysis of eye movements in a simple list-reading task. The strength of the computationally rational approach is in the…
(more)
▼ This dissertation pursues a computationally rational analysis of eye movements in a simple list-reading task. The strength of the computationally rational approach is in the ability to explain why certain phenomena may emerge under the assumption that behavior is an approximately optimal adaptation to the joint constraints of an organism's intrinsic computational constraints and task demands. The provided theory and model integrates a framework of lexical processing as active perception (Norris, 2006) with oculomotor constraints derived from a broad-coverage model of eye movement control in reading (Reichle, Warren & McConnell 2009). The first portion of the thesis provides experimental evidence of adaptation of fixation durations to quantitatively-expressed payoffs in a simple reading task, and adaptation in the model on the same dimension. The second portion explores spillover lexical frequency effects in the same framework and how they may emerge from a model that can adaptively allocate processing resources to information drawn from perception (foveal or parafoveal), or memory. In addition to implications for eye movement control in reading, these findings can be interpreted to bear on task adaptation in reading, as well as the adaptive use of perception and memory in a sequential sampling framework.
Advisors/Committee Members: Lewis, Richard L. (committee member), Baveja, Satinder Singh (committee member), Brennan, Jonathan R. (committee member), Boland, Julie (committee member), Hale, John T. (committee member).
Subjects/Keywords: eye movements; eyetracking; computationally rational analysis; computational modeling; lexical decision; sequential sampling; Psychology; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Shvartsman, M. (2014). Adaptive Eye Movement Control in a Simple Linguistic Task. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/110380
Chicago Manual of Style (16th Edition):
Shvartsman, Michael. “Adaptive Eye Movement Control in a Simple Linguistic Task.” 2014. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/110380.
MLA Handbook (7th Edition):
Shvartsman, Michael. “Adaptive Eye Movement Control in a Simple Linguistic Task.” 2014. Web. 04 Mar 2021.
Vancouver:
Shvartsman M. Adaptive Eye Movement Control in a Simple Linguistic Task. [Internet] [Doctoral dissertation]. University of Michigan; 2014. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/110380.
Council of Science Editors:
Shvartsman M. Adaptive Eye Movement Control in a Simple Linguistic Task. [Doctoral Dissertation]. University of Michigan; 2014. Available from: http://hdl.handle.net/2027.42/110380
26.
Newman, Lee I.
Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task.
Degree: PhD, Computer Science and Engineering and Psychology, 2009, University of Michigan
URL: http://hdl.handle.net/2027.42/64780
► Decision making, the process of choosing among a set of options, is a fundamental aspect of everyday mental life. Decisions are often made under conditions…
(more)
▼ Decision making, the process of choosing among a set of options, is a fundamental aspect of everyday mental life. Decisions are often made under conditions of uncertainty, when the payoffs are probabilistic and unknown. Two important challenges in the study of decision making are to understand how decision making processes are instantiated computationally and to reveal and characterize differences in decision making processes across individuals. In this dissertation I used computational and behavioral methods to study decision making under uncertainty in the context of the Iowa Gambling Task (IGT). The main contributions are: (i) A biologically-grounded computational model that provides a better account of IGT behavior than the most widely accepted model; (ii) The identification of three fundamentally different decision making styles in the IGT; (iii) An improved conceptualization of decision making that offers a more comprehensive approach to analyzing performance and that has important implications for the study of normal and clinically impaired decision making; (iv) An empirical challenge to the widely held belief that IGT performance is associated with impulsive and risky decision making; (v) A demonstrated association between decision making in the IGT and cognitive abilities as measured by the Wisconsin Card Sorting Task; and (vi) the introduction and successful demonstration of a robust data clustering methodology that offers significant advantages over methods that are currently used in the field of Psychology.
Advisors/Committee Members: Polk, Thad A. (committee member), Baveja, Satinder Singh (committee member), Laird, John E. (committee member), Lewis, Richard L. (committee member), Meyer, David E. (committee member).
Subjects/Keywords: Judgment and Decision Making Under Uncertainty; Iowa Gambling Task; Computational and Cognitive Modeling, Reinforcement Learning; Choice Behavior, Probabalistic Choice, Impaired Decision Making; Ensemble Clustering; Prefrontal Cortex and Cognitive Control; Computer Science; Psychology; Engineering; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Newman, L. I. (2009). Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/64780
Chicago Manual of Style (16th Edition):
Newman, Lee I. “Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task.” 2009. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/64780.
MLA Handbook (7th Edition):
Newman, Lee I. “Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task.” 2009. Web. 04 Mar 2021.
Vancouver:
Newman LI. Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task. [Internet] [Doctoral dissertation]. University of Michigan; 2009. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/64780.
Council of Science Editors:
Newman LI. Decision Making under Uncertainty: Revealing, Characterizing and Modeling Individual Differences in the Iowa Gambling Task. [Doctoral Dissertation]. University of Michigan; 2009. Available from: http://hdl.handle.net/2027.42/64780
27.
Li, Ning Hui.
The Goal Re-activation Problem in Cognitive Architectures.
Degree: PhD, Computer Science and Engineering, 2016, University of Michigan
URL: http://hdl.handle.net/2027.42/120676
► Intelligent agents in the real world have to manage multiple goals. However, the pursuit of some goals may only be possible under specific conditions which,…
(more)
▼ Intelligent agents in the real world have to manage multiple goals. However, the pursuit of some goals may only be possible under specific conditions which, if not met, requires the agent to suspend the goals for future re-activation. In cognitive architectures, suspended goals are stored in long-term memory; however, this prevents agents from automatically recognizing future opportunities to complete those goals.
This thesis characterizes this previously-unidentified problem as an instance of a more general circular dependency between the retrieval and use of knowledge in cognitive architectures: the agent must recognize that a goal is relevant to retrieve it, but cannot recognize it as such without retrieving it in the first place. We apply this characterization to develop preemptive and spontaneous retrieval strategies for goal re-activation in the Soar cognitive architecture. Evaluation of these strategies in an abstract domain shows that the spontaneous retrieval strategy dominates the other strategies, achieving higher goal completion rates in fewer operations, although it is also not without failure cases. Both types of strategies not only provide solutions to the goal re-activation problem, but also pave the way for further exploration of how intelligent agents can access the right knowledge at the right time.
Advisors/Committee Members: Laird, John E (committee member), Lewis, Richard L (committee member), Baveja, Satinder Singh (committee member), Durfee, Edmund H (committee member), Lebiere, Christian J. (committee member).
Subjects/Keywords: cognitive architecture; prospective memory; Computer Science; Engineering
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Li, N. H. (2016). The Goal Re-activation Problem in Cognitive Architectures. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/120676
Chicago Manual of Style (16th Edition):
Li, Ning Hui. “The Goal Re-activation Problem in Cognitive Architectures.” 2016. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/120676.
MLA Handbook (7th Edition):
Li, Ning Hui. “The Goal Re-activation Problem in Cognitive Architectures.” 2016. Web. 04 Mar 2021.
Vancouver:
Li NH. The Goal Re-activation Problem in Cognitive Architectures. [Internet] [Doctoral dissertation]. University of Michigan; 2016. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/120676.
Council of Science Editors:
Li NH. The Goal Re-activation Problem in Cognitive Architectures. [Doctoral Dissertation]. University of Michigan; 2016. Available from: http://hdl.handle.net/2027.42/120676
28.
Strohminger, Nina S.
The Hedonics of Disgust.
Degree: PhD, Psychology, 2013, University of Michigan
URL: http://hdl.handle.net/2027.42/97960
► Disgust is a negative emotion, and as such, it is frequently assumed that its only function is to generate negative evaluations. This dissertation aims to…
(more)
▼ Disgust is a negative emotion, and as such, it is frequently assumed that its only function is to generate negative evaluations. This dissertation aims to demonstrate that disgust can generate positive evaluations in the right context.
In Chapter I, we show how the existence of hedonic disgust would be both empirically novel and counter to predictions made by many current theories of emotion. We suggest that the impact of disgust depends on context, consistent with feelings-as- information theory (Schwarz, 2012). To this end, we lay out potential circumstances under which disgust may be experienced as enjoyable.
In Chapter II, we show that priming disgust with verbal stimuli leads people to rate both cartoons and moral violations as funnier, and food pictures as less appetizing. We also find that sad verbal stimuli enhance cartoon ratings, helping to rule out arousal level as the sole explanation for these effects.
In Chapter III, we show that disgusting verbal stimuli enhance enjoyment of abstract and grotesque art, but not landscape art. We further demonstrate that the effect of these disgust primes changes depending on the probe, with disgust increasing both the likability and offensiveness of judged paintings, while leading to lower ratings of prettiness.
In Chapter IV, we use two additional methods of inducing disgust—a noxious odor, and a filthy environment—to see whether this would enhance enjoyment of stand-up comedy and adventurous eating shows. We find that both of these disgust manipulations decrease enjoyment of traditional cooking shows, and increase enjoyment of adventurous eating shows. The experimental condition had no impact on humor judg- ments, counter to both feelings-as-information theory and congruency-based emotion theories.
In Chapter V, we discuss the scope and limitations of the current studies. While these studies demonstrate that disgust stimuli can have a positive effect on judgments, this effect is contingent on a variety of factors which are not yet fully understood. Furthermore, the precise mechanism by which incidental disgust leads to enjoyment remains unclear. Overall, this work shows that the influence of disgust is highly context-sensitive, and occasionally favorable, opening up a previously unexplored avenue for emotions research.
Advisors/Committee Members: Lewis, Richard L. (committee member), Schwarz, Norbert W. (committee member), Railton, Peter A. (committee member), Ellsworth, Phoebe C. (committee member), Meyer, David E. (committee member), Nesse, Randolph M. (committee member).
Subjects/Keywords: Disgust; Emotion; Benign Masochism; Judgment and Decision-making; Humor; Aesthetics; Psychology; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Strohminger, N. S. (2013). The Hedonics of Disgust. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/97960
Chicago Manual of Style (16th Edition):
Strohminger, Nina S. “The Hedonics of Disgust.” 2013. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/97960.
MLA Handbook (7th Edition):
Strohminger, Nina S. “The Hedonics of Disgust.” 2013. Web. 04 Mar 2021.
Vancouver:
Strohminger NS. The Hedonics of Disgust. [Internet] [Doctoral dissertation]. University of Michigan; 2013. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/97960.
Council of Science Editors:
Strohminger NS. The Hedonics of Disgust. [Doctoral Dissertation]. University of Michigan; 2013. Available from: http://hdl.handle.net/2027.42/97960
University of Michigan
29.
Kopecky, Jonathon Joseph.
Strategic Individual Differences in High-Speed Human Performance.
Degree: PhD, Psychology, 2008, University of Michigan
URL: http://hdl.handle.net/2027.42/61626
► Previous multitasking research has demonstrated a need to consider people's task-scheduling strategies when characterizing cognitive software and hardware limitations. The present dissertation reports three projects…
(more)
▼ Previous multitasking research has demonstrated a need to consider people's task-scheduling strategies when characterizing cognitive software and hardware limitations. The present dissertation reports three projects that, together, help illuminate the nature of such strategies.
The first project revealed that people tend to choose scheduling strategies that maximize their monetary reward for making precisely timed physical responses. Using a simple reaction-time (RT) procedure that required two key presses with rewards based solely on the inter-response interval (IRI) and order of responses, we found that participants adapted their RT and accuracy to the payoff. Participants who were given payoffs that emphasized shorter IRIs were both faster and less accurate than those who were given payoffs that emphasized accuracy. We formulated mathematical models of optimal strategy in terms of the external payoff and total level of internal and external system noise. Model-date fits indicated that people adapted near optimally to low and medium but not high levels of external noise.
The second project examined scheduling strategies for dealing with task interruptions. We found a significant slowdown in task interruptions when the interrupted and interrupting tasks had the same response modality, but that there was no such slowdown when the tasks had different response modalities. We developed computational models based on the CORE (Constraint-based Optimizing Reasoning Engine) theoretical framework to account for results from a representative task-interruption procedure. We found that motor preparation may begin ``locked'' for an interrupted task and ``unlocked'' only after a subsequent control signal is received from a process in the interrupting task.
The third project further investigated possible sources of individual differences in multitasking. Previous research has shown that collectivist cultures encourage ``holistic'' cognition strategies whereas individualistic cultures encourage ``analytic'' strategies. We reasoned that holistic cognition should aid people in performing two tasks simultaneously, whereas analytic cognition should aid people in switching between two tasks. Evidence from time-sharing and task-switching experiments supported this hypothesis. Japanese participants preferred simultaneous multitasking, whereas American participants preferred sequential multitasking.
Our research provides important techniques to disentangle cognitive structural limitations from software limitations for explaining people's strategic choices.
Advisors/Committee Members: Lewis, Richard L. (committee member), Meyer, David E. (committee member), Kitayama, Shinobu (committee member), Liu, Yili (committee member).
Subjects/Keywords: Multitasking; Task Interruptions; Cultural Differences; CORE Computational Modeling; Speed-Accuracy Tradeoff; Science; Social Sciences
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Kopecky, J. J. (2008). Strategic Individual Differences in High-Speed Human Performance. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/61626
Chicago Manual of Style (16th Edition):
Kopecky, Jonathon Joseph. “Strategic Individual Differences in High-Speed Human Performance.” 2008. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/61626.
MLA Handbook (7th Edition):
Kopecky, Jonathon Joseph. “Strategic Individual Differences in High-Speed Human Performance.” 2008. Web. 04 Mar 2021.
Vancouver:
Kopecky JJ. Strategic Individual Differences in High-Speed Human Performance. [Internet] [Doctoral dissertation]. University of Michigan; 2008. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/61626.
Council of Science Editors:
Kopecky JJ. Strategic Individual Differences in High-Speed Human Performance. [Doctoral Dissertation]. University of Michigan; 2008. Available from: http://hdl.handle.net/2027.42/61626
University of Michigan
30.
Wintermute, Samuel B.
Abstraction, Imagery, and Control in Cognitive Architecture.
Degree: PhD, Computer Science & Engineering, 2010, University of Michigan
URL: http://hdl.handle.net/2027.42/78795
► This dissertation presents a theory describing the components of a cognitive architecture supporting intelligent behavior in spatial tasks. In this theory, an abstract symbolic representation…
(more)
▼ This dissertation presents a theory describing the components of a cognitive architecture supporting intelligent behavior in spatial tasks. In this theory, an abstract symbolic representation serves as the basis for decisions. As a means to support abstract decision-making, imagery processes are also present. Here, a concrete (highly detailed) representation of the state of the problem is maintained in parallel with the abstract representation. Perceptual and action systems are decomposed into parts that operate between the environment and the concrete representation, and parts that operate between the concrete and abstract representations. Control processes can issue actions as a continuous function of information in the concrete representation, and actions can be simulated (imagined) in terms of it. The agent can then derive useful abstract information by applying perceptual processes to the resulting concrete state.
This theory addresses two challenges in architecture design that arise due to the diversity and complexity of spatial tasks that an intelligent agent must address. The perceptual abstraction problem results from the difficulty of creating a single perception system able to induce appropriate abstract representations in each of the many tasks an agent might encounter, and the irreducibility problem arises because some tasks are resistant to being abstracted at all. Imagery works to mitigate the perceptual abstraction problem by allowing a given perception system to work in more tasks, as perception can be dynamically combined with imagery. Continuous control, and the simulation thereof via imagery, works to mitigate the irreducibility problem. The use of imagery to address these challenges differs from other approaches in AI, where imagery is considered as an alternative to abstract representation, rather than as a means to it.
A detailed implementation of the theory is described, which is an extension of the Soar cognitive architecture. Agents instantiated in this architecture are demonstrated, including agents that use reinforcement learning and imagery to play arcade games, and an agent that performs sampling-based motion planning for a car-like vehicle. The performance of these agents is discussed in the context of the underlying architectural theory. Connections between this work and psychological theories of mental imagery are also discussed.
Advisors/Committee Members: Laird, John E. (committee member), Baveja, Satinder Singh (committee member), Kuipers, Benjamin (committee member), Lewis, Richard L. (committee member).
Subjects/Keywords: Cognitive Architecture; Artificial Intelligence; Mental Imagery; Computer Science; Engineering; Science
Record Details
Similar Records
Cite
Share »
Record Details
Similar Records
Cite
« Share
❌
APA ·
Chicago ·
MLA ·
Vancouver ·
CSE |
Export
to Zotero / EndNote / Reference
Manager
APA (6th Edition):
Wintermute, S. B. (2010). Abstraction, Imagery, and Control in Cognitive Architecture. (Doctoral Dissertation). University of Michigan. Retrieved from http://hdl.handle.net/2027.42/78795
Chicago Manual of Style (16th Edition):
Wintermute, Samuel B. “Abstraction, Imagery, and Control in Cognitive Architecture.” 2010. Doctoral Dissertation, University of Michigan. Accessed March 04, 2021.
http://hdl.handle.net/2027.42/78795.
MLA Handbook (7th Edition):
Wintermute, Samuel B. “Abstraction, Imagery, and Control in Cognitive Architecture.” 2010. Web. 04 Mar 2021.
Vancouver:
Wintermute SB. Abstraction, Imagery, and Control in Cognitive Architecture. [Internet] [Doctoral dissertation]. University of Michigan; 2010. [cited 2021 Mar 04].
Available from: http://hdl.handle.net/2027.42/78795.
Council of Science Editors:
Wintermute SB. Abstraction, Imagery, and Control in Cognitive Architecture. [Doctoral Dissertation]. University of Michigan; 2010. Available from: http://hdl.handle.net/2027.42/78795
◁ [1] [2] ▶
. 
Open Access Theses and Dissertations
