You are in: Submissions > Select unit of assessment > UOA 23 Computer Science and Informatics > University of Bristol > RA1, RA2 and RA5c

UOA 23 - Computer Science and Informatics

University of Bristol

RA1, RA2 and RA5c: Staff and output details and Category C staff circumstances

 

Bogacz, R - Category : A

Research Groups:

A - Intelligent Systems

RA2 - Research outputs:

Number of outputs: 4

Output number: 1 of 4

Title Model of familiarity discrimination in the perirhinal cortex
Output type: Journal article
Journal title: Journal of Computational Neuroscience
Month/year of publication: January 2001
Pagination: 5-23
Volume: 10 (1)
ISSN: 09295313
DOI: 10.1023/A:1008925909305 ?
Interdisciplinary output: Yes
Research group: A - Intelligent Systems
Co-authors: Brown, MW, Giraud-Carrier, CG
Other relevant details: This paper uses techniques from computer science and information theory to answer a fundamental question regarding the memory system in the brain: How can the brain region dealing with familiarity discrimination (called the perirhinal cortex) produce powerful human familiarity discrimination abilities? The paper shows formally and for the first time that, since familiarity discrimination is computationally easier than recollection, the neural networks of the perirhinal cortex can discriminate familiarity for many more stimuli than all other remaining memory areas can recollect. Journal of Computational Neuroscience has 2004 impact factor of 2.2.

Output number: 2 of 4

Title Comparison of computational models of familiarity discrimination in the perirhinal cortex
Output type: Journal article
Journal title: Hippocampus
Month/year of publication: April 2003
Pagination: 494-524
Volume: 13 (4)
ISSN: 10509631
URL: Original article on web ?
DOI: 10.1002/hipo.10093 ?
Interdisciplinary output: Yes
Research group: A - Intelligent Systems
Co-authors: Brown, MW
Other relevant details: This paper analyses and compares five published computational models of familiarity discrimination in the perirhinal cortex. It shows that only one of them can perform familiarity discrimination as efficiently as humans do, while being consistent with neurobiological data. This is important because it provides constraints any such model should satisfy. This has been used in subsequent work by Norman (Princeton) to propose a new model for familiarity discrimination that satisfies these constraints. Hippocampus has 2004 impact factor of 4.5.

Output number: 3 of 4

Title The physics of optimal decision making: a formal analysis of models of performance in two-alternative forced- choice tasks
Output type: Journal article
Journal title: Psychological Review
Month/year of publication: October 2006
Pagination: 700-765
Volume: 113 (4)
ISSN: 0033295X
DOI: 10.1037/0033-295X.113.4.700 ?
Interdisciplinary output: Yes
Research group: A - Intelligent Systems
Co-authors: Brown, E, Holmes, P, Moehlis, J
Number of additional co-authors: 1
Other relevant details: This 66-page paper provides an extensive analysis of five computational models of decision making in the brain, as opposed to familiarity discrimination which was covered in the previous paper. The paper finds constraints they need to satisfy to implement a statistically optimal decision making algorithm, and shows that, for optimal parameters, these models become computationally equivalent. It makes a range of experimental predictions that have subsequently been verified. The work has been continued by groups in Princeton and Technion (Israel). Psychological Review had an impact factor of 7.1 in 2004, and is the most prestigious journal in psychology.

Output number: 4 of 4

Title The basal ganglia and cortex implement optimal decision making between alternative actions
Output type: Journal article
Journal title: Neural Computation
Month/year of publication: February 2007
Pagination: 442-477
Volume: 19 (2)
ISSN: 08997667
URL: Original article on web ?
DOI: 10.1162/neco.2007.19.2.442 ?
Interdisciplinary output: Yes
Research group: A - Intelligent Systems
Co-authors: Gurney, K
Other relevant details: This paper proposes that neural networks in a part of the brain called the basal ganglia implement a statistically optimal decision making algorithm MSPRT. It shows that the equation describing MSPRT can be mapped onto the anatomy of the basal ganglia, thus it provides a rationale for the observed organization, which used to be an open problem. The mapping between MSPRT and the anatomy predicts that one of the basal nuclei should compute an exponent of its input, and this prediction has been verified in existing experimental data. Neural Computation has 2004 impact factor 2.4, which is the highest in computational neuroscience.