Monte Carlo Functional Regularisation for Continual Learning

TL;DR

This paper introduces MCFRCL, a novel continual learning framework that uses Monte Carlo sampling and distribution-based regularisation to improve prediction accuracy and efficiency in neural networks.

Contribution

The paper proposes a new functional regularisation method for continual learning that leverages Monte Carlo sampling and distribution metrics to reduce computational costs and approximation errors.

Findings

MCFRCL outperforms existing methods on MNIST and CIFAR datasets.

It achieves higher prediction accuracy with lower computational costs.

The approach effectively captures statistical characteristics of model predictions.

Abstract

Continual learning (CL) is crucial for the adaptation of neural network models to new environments. Although outperforming weight-space regularisation approaches, the functional regularisation-based CL methods suffer from high computational costs and large linear approximation errors. In this work, we present a new functional regularisation CL framework, called MCFRCL, which approximates model prediction distributions by Monte Carlo (MC) sampling. Moreover, three continuous distributions are leveraged to capture the statistical characteristics of the MC samples via moment-based methods. Additionally, both the Wasserstein distance and the Kullback-Leibler (KL) distance are employed to construct the regularisation function. The proposed MCFRCL is evaluated against multiple benchmark methods on the MNIST and CIFAR datasets, with simulation results highlighting its effectiveness in both prediction accuracy and training efficiency.