Conjunctive coding in an evolved spiking model of retrosplenial cortex

Retrosplenial cortex (RSC) is an association cortex supporting spatial navigation and memory. However, critical issues remain concerning the forms by which its ensemble spiking patterns register spatial relationships that are difficult for experimental techniques to fully address. We therefore appli...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Behavioral neuroscience Ročník 132; číslo 5; s. 430
Hlavní autori: Rounds, Emily L, Alexander, Andrew S, Nitz, Douglas A, Krichmar, Jeffrey L
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: United States 01.10.2018
Predmet:
Action Potentials
Algorithms
Animals
Cerebral Cortex - cytology
Cerebral Cortex - physiology
Male
Models, Neurological
Neurons - cytology
Neurons - physiology
Rats, Long-Evans
Spatial Navigation - physiology
ISSN:1939-0084, 1939-0084
On-line prístup:Zistit podrobnosti o prístupe
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Retrosplenial cortex (RSC) is an association cortex supporting spatial navigation and memory. However, critical issues remain concerning the forms by which its ensemble spiking patterns register spatial relationships that are difficult for experimental techniques to fully address. We therefore applied an evolutionary algorithmic optimization technique to create spiking neural network models that matched electrophysiologically observed spiking dynamics in rat RSC neuronal ensembles. Virtual experiments conducted on the evolved networks revealed a mixed selectivity coding capability that was not built into the optimization method, but instead emerged as a consequence of replicating biological firing patterns. The experiments reveal several important outcomes of mixed selectivity that may subserve flexible navigation and spatial representation: (a) robustness to loss of specific inputs, (b) immediate and stable encoding of novel routes and route locations, (c) automatic resolution of input variable conflicts, and (d) dynamic coding that allows rapid adaptation to changing task demands without retraining. These findings suggest that biological retrosplenial cortex can generate unique, first-trial, conjunctive encodings of spatial positions and actions that can be used by downstream brain regions for navigation and path integration. Moreover, these results are consistent with the proposed role for the RSC in the transformation of representations between reference frames and navigation strategy deployment. Finally, the specific modeling framework used for evolving synthetic retrosplenial networks represents an important advance for computational modeling by which synthetic neural networks can encapsulate, describe, and predict the behavior of neural circuits at multiple levels of function. (PsycINFO Database Record (c) 2018 APA, all rights reserved).
Bibliografia:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:1939-0084
1939-0084
DOI:10.1037/bne0000236