Emergent mechanisms for long timescales depend on training curriculum and affect performance in memory tasks

TL;DR

This study investigates how training curriculum influences the development of long timescales in recurrent neural networks, revealing different mechanisms for memory retention and improved task performance.

Contribution

It demonstrates that training curriculum determines whether RNNs develop longer neuron timescales or rely on recurrent connectivity, enhancing memory task performance and generalization.

Findings

Multi-head curriculum maintains constant neuron timescales while developing longer network timescales.

Multi-head training improves speed, stability, and generalization of RNNs on memory tasks.

Adapting timescales via recurrent interactions enhances learning of complex temporal objectives.

Abstract

Recurrent neural networks (RNNs) in the brain and in silico excel at solving tasks with intricate temporal dependencies. Long timescales required for solving such tasks can arise from properties of individual neurons (single-neuron timescale, $\tau$, e.g., membrane time constant in biological neurons) or recurrent interactions among them (network-mediated timescale). However, the contribution of each mechanism for optimally solving memory-dependent tasks remains poorly understood. Here, we train RNNs to solve $N$-parity and $N$-delayed match-to-sample tasks with increasing memory requirements controlled by $N$ by simultaneously optimizing recurrent weights and $\tau$s. We find that for both tasks RNNs develop longer timescales with increasing $N$, but depending on the learning objective, they use different mechanisms. Two distinct curricula define learning objectives: sequential learning of a single-$N$ (single-head) or simultaneous learning of multiple $N$s (multi-head). Single-head networks increase their $\tau$ with $N$ and are able to solve tasks for large $N$, but they suffer from catastrophic forgetting. However, multi-head networks, which are explicitly required to hold multiple concurrent memories, keep $\tau$ constant and develop longer timescales through recurrent connectivity. Moreover, we show that the multi-head curriculum increases training speed and network stability to ablations and perturbations, and allows RNNs to generalize better to tasks beyond their training regime. This curriculum also significantly improves training GRUs and LSTMs for large-$N$ tasks. Our results suggest that adapting timescales to task requirements via recurrent interactions allows learning more complex objectives and improves the RNN's performance.