Modelling continual learning in humans with Hebbian context gating and exponentially decaying task signals

TL;DR

This paper introduces a biologically inspired neural network model with Hebbian gating and task signals that effectively learns multiple tasks sequentially without forgetting, aligning with human learning patterns.

Contribution

The paper presents a novel neural network architecture with 'sluggish' units and Hebbian learning to mitigate interference in continual learning, inspired by primate prefrontal cortex mechanisms.

Findings

Model matches human performance on blocked and interleaved training.

Gating scheme creates orthogonal, interference-resistant representations.

Performance differences linked to misestimation of category boundaries.

Abstract

Humans can learn several tasks in succession with minimal mutual interference but perform more poorly when trained on multiple tasks at once. The opposite is true for standard deep neural networks. Here, we propose novel computational constraints for artificial neural networks, inspired by earlier work on gating in the primate prefrontal cortex, that capture the cost of interleaved training and allow the network to learn two tasks in sequence without forgetting. We augment standard stochastic gradient descent with two algorithmic motifs, so-called "sluggish" task units and a Hebbian training step that strengthens connections between task units and hidden units that encode task-relevant information. We found that the "sluggish" units introduce a switch-cost during training, which biases representations under interleaved training towards a joint representation that ignores the contextual cue, while the Hebbian step promotes the formation of a gating scheme from task units to the hidden layer that produces orthogonal representations which are perfectly guarded against interference. Validating the model on previously published human behavioural data revealed that it matches performance of participants who had been trained on blocked or interleaved curricula, and that these performance differences were driven by misestimation of the true category boundary.