國立虎尾科技大學 |

Distributed Representations and Localized Computations in the Ventral Visual Cortex.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Distributed Representations and Localized Computations in the Ventral Visual Cortex./
Author:	Shehzad, Zarrar.
Description:	1 online resource (274 pages)
Notes:	Source: Dissertation Abstracts International, Volume: 79-05(E), Section: B.
Contained By:	Dissertation Abstracts International79-05B(E).
Subject:	Psychology. -
Online resource:	click for full text (PQDT)
ISBN:	9780355681994

Distributed Representations and Localized Computations in the Ventral Visual Cortex.
Shehzad, Zarrar.

Distributed Representations and Localized Computations in the Ventral Visual Cortex. - 1 online resource (274 pages)

Source: Dissertation Abstracts International, Volume: 79-05(E), Section: B.

Thesis (Ph.D.)--Yale University, 2017.

Includes bibliographical references

The recognition of faces includes many processes from low-level analysis of face structure to person knowledge. Two perspectives of brain organization for face processing have been widely discussed. One perspective, supported by brain lesion and activation studies, proposes that domain-specific information is represented in localized regions. A second perspective, supported by multi-voxel pattern analysis (MVPA) of brain imaging data, argues that it would be inefficient to have separate domain-specific regions for every possible function. Consequently, information must be distributed and the unique combination of activity across regions will encode a particular representation. Here we consider a third perspective and propose that domain-specific computations are localized but representations are distributed. Provided that each region of the face network can accept inputs from any region, information would need to be shared between regions for domain-specific computations to occur in localized regions. The distributed nature of representations creates ambiguity about the source of variation in the activity of a region. We can better characterize face representations by removing shared information from other regions and using behavioral and computational measures to identify the sources of variation within each face-selective region.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2018

Mode of access: World Wide Web

ISBN: 9780355681994Subjects--Topical Terms:

555998
Psychology.
Index Terms--Genre/Form:

554714
Electronic books.

Distributed Representations and Localized Computations in the Ventral Visual Cortex.
LDR:04412ntm a2200349Ki 4500 001 918549
005 20181026115418.5
006 m o u
007 cr mn||||a|a||
008 190606s2017 xx obm 000 0 eng d
020 $a 9780355681994
035 $a (MiAaPQ)AAI10767285
035 $a AAI10767285
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Shehzad, Zarrar. $3 1192897
245 1 0 $a Distributed Representations and Localized Computations in the Ventral Visual Cortex.
264 0 $c 2017
300 $a 1 online resource (274 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertation Abstracts International, Volume: 79-05(E), Section: B.
500 $a Adviser: Gregory McCarthy.
502 $a Thesis (Ph.D.)--Yale University, 2017.
504 $a Includes bibliographical references
520 $a The recognition of faces includes many processes from low-level analysis of face structure to person knowledge. Two perspectives of brain organization for face processing have been widely discussed. One perspective, supported by brain lesion and activation studies, proposes that domain-specific information is represented in localized regions. A second perspective, supported by multi-voxel pattern analysis (MVPA) of brain imaging data, argues that it would be inefficient to have separate domain-specific regions for every possible function. Consequently, information must be distributed and the unique combination of activity across regions will encode a particular representation. Here we consider a third perspective and propose that domain-specific computations are localized but representations are distributed. Provided that each region of the face network can accept inputs from any region, information would need to be shared between regions for domain-specific computations to occur in localized regions. The distributed nature of representations creates ambiguity about the source of variation in the activity of a region. We can better characterize face representations by removing shared information from other regions and using behavioral and computational measures to identify the sources of variation within each face-selective region.
520 $a In each of the three studies of this dissertation, we find distributed representations for faces and person knowledge. However, in each study, we find evidence that the underlying computations were localized. Study 1 revealed that category representations are localized for faces, vehicles, letters, and objects when we removed shared information by combining activity patterns from different regions into one model. Information was being redundantly shared. For instance, the same activity patterns for face stimuli were found in the fusiform face area (FFA) and other regions but only unique information for faces was localized to the FFA. Study 2 revealed that person knowledge for faces was localized in the left ventral anterior temporal lobe (vATL) and feedback occurred from the vATL to FFA. This feedback may explain distributed representations for person knowledge in both vATL and FFA. Finally, Study 3 revealed that while face activity was distributed, specific computations were localized between hemispheres. Within face-selective regions, the right FFA processed face-likeness, activating to a face that looked more like the average of all faces, and the left FFA/anterior fusiform (aFus) processed person-likeness, activating to a face that looked unlike a person's average face.
520 $a The results of this dissertation demonstrate that representations are distributed but computations are localized within the face network. The right FFA is selective for faces in general, the left FFA/aFus is selective for person identity and within-person variability, and the left vATL is selective for person knowledge. Information between these and other regions is shared acting as input to other regions of the face network. The redundant sharing of information may allow the same representations to be seen from many perspectives.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2018
538 $a Mode of access: World Wide Web
650 4 $a Psychology. $3 555998
650 4 $a Neurosciences. $3 593561
655 7 $a Electronic books. $2 local $3 554714
690 $a 0621
690 $a 0317
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Yale University. $3 1178968
773 0 $t Dissertation Abstracts International $g 79-05B(E).
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10767285 $z click for full text (PQDT)