Recurrent Network Models Of Sequence Generation And Memory

TL;DR

This paper shows that largely unstructured recurrent neural networks can learn to generate sequences and perform working memory tasks efficiently by modifying only a small fraction of connections, without specialized architecture.

Contribution

It introduces Partial In-Network Training (PINning), demonstrating that disordered networks can produce sequences and memory, matching experimental data without pre-wired mechanisms.

Findings

Disordered networks can learn sequences with minimal connection modifications.

Sequences emerge from cooperation between recurrent interactions and external inputs.

Model matches cellular imaging data from posterior parietal cortex during memory tasks.

Abstract

Sequential activation of neurons is a common feature of network activity during a variety of behaviors, including working memory and decision making. Previous network models for sequences and memory emphasized specialized architectures in which a principled mechanism is pre-wired into their connectivity. Here we demonstrate that, starting from random connectivity and modifying a small fraction of connections, a largely disordered recur- rent network can produce sequences and implement working memory efficiently. We use this process, called Partial In-Network Training (PINning), to model and match cellular resolution imaging data from the posterior parietal cortex during a virtual memory- guided two-alternative forced-choice task. Analysis of the connectivity reveals that sequences propagate by the cooperation between recurrent synaptic interactions and external inputs, rather than through feedforward or asymmetric connections. Together our results suggest that neural sequences may emerge through learning from largely unstructured network architectures.