Dynamical phases of short-term memory mechanisms in RNNs

TL;DR

This paper explores neural mechanisms of short-term memory in RNNs, introducing two models supported by analytical and empirical analysis, and provides theoretical scaling laws and a large dataset for further research.

Contribution

It introduces two novel mechanisms for short-term memory in RNNs, supported by analytical models and extensive empirical validation, advancing understanding of neural memory processes.

Findings

Two mechanisms support short-term memory: slow-point manifolds and limit cycles.

Derived scaling laws for critical learning rates related to delay periods.

Evaluated approximately 80,000 RNNs to validate theoretical predictions.

Abstract

Short-term memory is essential for cognitive processing, yet our understanding of its neural mechanisms remains unclear. Neuroscience has long focused on how sequential activity patterns, where neurons fire one after another within large networks, can explain how information is maintained. While recurrent connections were shown to drive sequential dynamics, a mechanistic understanding of this process still remains unknown. In this work, we introduce two unique mechanisms that can support this form of short-term memory: slow-point manifolds generating direct sequences or limit cycles providing temporally localized approximations. Using analytical models, we identify fundamental properties that govern the selection of each mechanism. Precisely, on short-term memory tasks (delayed cue-discrimination tasks), we derive theoretical scaling laws for critical learning rates as a function of the delay period length, beyond which no learning is possible. We empirically verify these results by training and evaluating approximately 80,000 recurrent neural networks (RNNs), which are publicly available for further analysis. Overall, our work provides new insights into short-term memory mechanisms and proposes experimentally testable predictions for systems neuroscience.