Neural representational geometries reflect behavioral differences in monkeys and recurrent neural networks

This study investigates how neural representational geometries reflect behavioral differences in monkeys performing a visually cued rule-based task. While both monkeys exhibited similar overall task performance, analysis of dorsolateral prefrontal cortex (PFdl) neural activity revealed striking differences in representational geometry. These differences correlated with subtle reaction time variations, suggesting the monkeys employed different strategies. Recurrent neural network (RNN) models, trained on the same task, demonstrated a similar relationship between representational geometry, training duration, and reaction times, providing a potential mechanistic explanation for the observed differences in the monkeys.

Valeria Fascianelli, Aldo Battista, Fabio Stefanini, Satoshi Tsujimoto, Aldo Genovesio, Stefano Fusi