Reusable Modular Architecture Enables Flexible Cognitive Operations in the Mouse Brain and Artificial Recurrent Networks

Complex behaviors are thought to be built by combining simpler cognitive components. Computational modeling has shown that artificial neural networks can perform a variety of tasks by flexibly combining functional modules, each specialized for a specific computation, to construct a complex task. However, it is unknown whether reusable modular networks are found in the brain. Here, we show that mice performing a delayed match-to-sample with delayed report (DMS-dr) task reuse neuronal subspaces that were specialized for stimulus processing and memory maintenance. These subspaces were reused during the task to represent new stimulus inputs and different types of memories, respectively. Clustering analyses showed each subspace was supported by a distinct cluster of neurons in prefrontal cortex and parietal cortex. Studying artificial recurrent networks constrained to neural data found silencing specific clusters disrupted specific computations, consistent with a modular and reusable organization. Altogether, our findings show that the brain can flexibly reuse computational components to perform a complex cognitive task.