東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Theoretical Investigations in Visual...

Orhan, A. Emin.

Linked to FindBook

Google Book

Amazon

博客來

Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding./
Author:	Orhan, A. Emin.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2013,
Description:	292 p.
Notes:	Source: Dissertation Abstracts International, Volume: 75-02(E), Section: B.
Contained By:	Dissertation Abstracts International75-02B(E).
Subject:	Experimental psychology. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3600443
ISBN:	9781303505485

Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding.
Orhan, A. Emin.

Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding. - Ann Arbor : ProQuest Dissertations & Theses, 2013 - 292 p.

Source: Dissertation Abstracts International, Volume: 75-02(E), Section: B.

Thesis (Ph.D.)--University of Rochester, 2013.

This item is not available from ProQuest Dissertations & Theses.

Visual short-term memory (VSTM) makes a stable and coherent perception of the world possible by maintaining visual information about a scene as it was a moment ago. As VSTM is widely considered to be a severely limited form of memory, it is important to know what kinds of information about a scene can or cannot be held in VSTM. This thesis proposes a probabilistic clustering model that attempts to capture the representations that subjects encode from brief presentations of simple visual scenes typically used in VSTM studies (Chapter 2). The model explains a number of previously observed biases in VSTM and predicts dependencies between representations of different stimuli that should decrease with the difference between their feature values. We show that results from three experiments are qualitatively in agreement with this novel prediction of the model. We further show that this type of dependency can arise as a natural consequence of the simultaneous encoding of multiple stimuli in a population of neurons with correlated responses (Chapter 4). We also investigate the consequences of possible mismatches, of the type demonstrated in Chapter 2, between the internal model that subjects might use to encode simple visual displays and the model actually used by the experimenter to generate the stimuli (Chapter 3). We tentatively suggest that such mismatches can arise due to the long-term adaptation of the visual system to the stimulus statistics in the natural environment which is generally quite different from the stimulus statistics used in VSTM studies. Remarkably, we show that this type of mismatch alone, without assuming any resource or capacity limitation, can explain some of the main qualitative characteristics of performance limitations observed in VSTM studies typically attributed to capacity limitations. Finally, we take a critical look at the use of impoverished visual stimuli that lack the rich statistical structure displayed by stimuli that are more representative of the natural environment and discuss how the study of VSTM can be extended to ecologically more realistic stimuli, as well as possible implications of such an extension (Chapter 5). Taken together, this thesis proposes novel theoretical ideas at multiple levels of analysis, from computation to neural implementation, for an ecologically more realistic and behaviorally more adequate modeling of representations in VSTM.

ISBN: 9781303505485Subjects--Topical Terms:

2144733
Experimental psychology.

Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding.
LDR:03525nmm a2200313 4500 001 2159204
005 20180622095236.5
008 190424s2013 ||||||||||||||||| ||eng d
020 $a 9781303505485
035 $a (MiAaPQ)AAI3600443
035 $a (MiAaPQ)rochester:10565
035 $a AAI3600443
040 $a MiAaPQ $c MiAaPQ
100 1 $a Orhan, A. Emin. $3 3347069
245 1 0 $a Theoretical Investigations in Visual Short-Term Memory: Structured Probabilistic Representations, Model Mismatch and Neural Population Coding.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2013
300 $a 292 p.
500 $a Source: Dissertation Abstracts International, Volume: 75-02(E), Section: B.
500 $a Adviser: Robert A. Jacobs.
502 $a Thesis (Ph.D.)--University of Rochester, 2013.
506 $a This item is not available from ProQuest Dissertations & Theses.
520 $a Visual short-term memory (VSTM) makes a stable and coherent perception of the world possible by maintaining visual information about a scene as it was a moment ago. As VSTM is widely considered to be a severely limited form of memory, it is important to know what kinds of information about a scene can or cannot be held in VSTM. This thesis proposes a probabilistic clustering model that attempts to capture the representations that subjects encode from brief presentations of simple visual scenes typically used in VSTM studies (Chapter 2). The model explains a number of previously observed biases in VSTM and predicts dependencies between representations of different stimuli that should decrease with the difference between their feature values. We show that results from three experiments are qualitatively in agreement with this novel prediction of the model. We further show that this type of dependency can arise as a natural consequence of the simultaneous encoding of multiple stimuli in a population of neurons with correlated responses (Chapter 4). We also investigate the consequences of possible mismatches, of the type demonstrated in Chapter 2, between the internal model that subjects might use to encode simple visual displays and the model actually used by the experimenter to generate the stimuli (Chapter 3). We tentatively suggest that such mismatches can arise due to the long-term adaptation of the visual system to the stimulus statistics in the natural environment which is generally quite different from the stimulus statistics used in VSTM studies. Remarkably, we show that this type of mismatch alone, without assuming any resource or capacity limitation, can explain some of the main qualitative characteristics of performance limitations observed in VSTM studies typically attributed to capacity limitations. Finally, we take a critical look at the use of impoverished visual stimuli that lack the rich statistical structure displayed by stimuli that are more representative of the natural environment and discuss how the study of VSTM can be extended to ecologically more realistic stimuli, as well as possible implications of such an extension (Chapter 5). Taken together, this thesis proposes novel theoretical ideas at multiple levels of analysis, from computation to neural implementation, for an ecologically more realistic and behaviorally more adequate modeling of representations in VSTM.
590 $a School code: 0188.
650 4 $a Experimental psychology. $3 2144733
650 4 $a Cognitive psychology. $3 523881
690 $a 0623
690 $a 0633
710 2 $a University of Rochester. $b School of Arts and Sciences. $3 2099014
773 0 $t Dissertation Abstracts International $g 75-02B(E).
790 $a 0188
791 $a Ph.D.
792 $a 2013
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3600443