Recurrent networks endowed with structural priors explain suboptimal animal behavior
Current Biology · 2023-02 · 33 (4) · 622-638.e7
DOI: 10.1016/j.cub.2022.12.044 ↗Kurz gefasst
Fachartikel in der Zeitschrift Current Biology, 2023-02. Verfasst von Manuel Molano-Mazón und 5 weiteren Autor:innen. Beteiligte Einrichtung: IDIBAPS, Rosselló 149, Barcelona 08036, Spain. Electronic address: manuelmolanomazon@gmail.com. Hauptthemen laut MEDLINE: Learning, Neural Networks, Computer. Die Arbeit wurde bisher 40-mal zitiert (Crossref) und nennt 100 Literaturangaben.
Abstract (Originalsprache)
The strategies found by animals facing a new task are determined both by individual experience and by structural priors evolved to leverage the statistics of natural environments. Rats quickly learn to capitalize on the trial sequence correlations of two-alternative forced choice (2AFC) tasks after correct trials but consistently deviate from optimal behavior after error trials. To understand this outcome-dependent gating, we first show that recurrent neural networks (RNNs) trained in the same 2AFC task outperform rats as they can readily learn to use across-trial information both after correct and error trials. We hypothesize that, although RNNs can optimize their behavior in the 2AFC task without any a priori restrictions, rats' strategy is constrained by a structural prior adapted to a natural environment in which rewarded and non-rewarded actions provide largely asymmetric information. When pre-training RNNs in a more ecological task with more than two possible choices, networks develop a strategy by which they gate off the across-trial evidence after errors, mimicking rats' behavior. Population analyses show that the pre-trained networks form an accurate representation of the sequence statistics independently of the outcome in the previous trial. After error trials, gating is implemented by a change in the network dynamics that temporarily decouple the categorization of the stimulus from the across-trial accumulated evidence. Our results suggest that the rats' suboptimal behavior reflects the influence of a structural prior that reacts to errors by isolating the network decision dynamics from the context, ultimately constraining the performance in a 2AFC laboratory task.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- IDIBAPS, Rosselló 149, Barcelona 08036, Spain. Electronic address: manuelmolanomazon@gmail.com.
- Laboratoire de Neurosciences Cognitives, INSERM U960, École Normale Supérieure - PSL Research University, 75005 Paris, France.
- IDIBAPS, Rosselló 149, Barcelona 08036, Spain.
- Center for Theoretical Neuroscience, Columbia University, 3227 Broadway, New York, NY 10027, USA; Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, 43 Vassar Street, Cambridge, MA 02139-…
Zitieren
APA
Molano-Mazón, M., Shao, Y., Duque, D., Yang, G. R., Ostojic, S., & de la Rocha, J. (2023). Recurrent networks endowed with structural priors explain suboptimal animal behavior. Current Biology, 33(4), 622-638.e7. https://doi.org/10.1016/j.cub.2022.12.044
Vancouver
Molano-Mazón M, Shao Y, Duque D, Yang GR, Ostojic S, de la Rocha J. Recurrent networks endowed with structural priors explain suboptimal animal behavior. Current Biology. 2023;33(4):622-638.e7. doi:10.1016/j.cub.2022.12.044
BibTeX
@article{MolanoMazn2023,
author = {Molano-Mazón, Manuel and Shao, Yuxiu and Duque, Daniel and Yang, Guangyu Robert and Ostojic, Srdjan and de la Rocha, Jaime},
title = {Recurrent networks endowed with structural priors explain suboptimal animal behavior},
journal = {Current Biology},
year = {2023},
volume = {33},
number = {4},
pages = {622--638.e7},
doi = {10.1016/j.cub.2022.12.044}
}
EUVIMED