Natural continual learning: success is a journey, not (just) a destination

By Ta-Chu Kao, Kristopher T. Jensen, Gido M. van de Ven, Alberto Bernacchia, Guillaume Hennequin. 2021

64 citations recorded by the source corpus.

Biological agents are known to learn many different tasks over the course of their lives, and to be able to revisit previous tasks and behaviors with little to no loss in performance. In contrast, artificial agents are prone to 'catastrophic forgetting' whereby performance on previous tasks deteriorates rapidly as new ones are acquired. This shortcoming has recently been addressed using methods that encourage parameters to stay close to those used for previous tasks. This can be done by (i) using specific parameter regularizers that map out suitable destinations in parameter space, or (ii) guiding the optimization journey by projecting gradients into subspaces that do not interfere with previous tasks. However, these methods often exhibit subpar performance in both feedforward and recurrent neural networks, with recurrent networks being of interest to the study of neural dynamics supporting biological continual learning. In this work, we propose Natural Continual Learning (NCL), a new method that unifies weight regularization and projected gradient descent. NCL uses Bayesian weight regularization to encourage good performance on all tasks at convergence and combines this with gradient projection using the prior precision, which prevents catastrophic forgetting during optimization. Our method outperforms both standard weight regularization techniques and projection based approaches when applied to continual learning problems in feedforward and recurrent networks. Finally, the trained networks evolve task-specific dynamics that are strongly preserved as new tasks are learned, similar to experimental findings in biological circuits.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics, Meta & Continual Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistical and Nonlinear Physics

In Meta & Continual Learning

Citations in the corpus

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, methodology): “1: k − 1 ) as a multivariate Gaussian with local gradient information [49], [50], [51], [52].”
Beyond Not-Forgetting: Continual Learning with Backward Knowledge Transfer (2022, methodology): “However, [10] leverages the knowledge of the current task in an implicit way to improve on the old tasks, whereas our method attempts more explicitly to update the model of old tasks whenever relevant new information is available.”
Energy-Based Models for Continual Learning (2020, methodology): “Regularization is used in CL to encourage the stability of those aspects of the network that are important for previous tasks (Li & Hoiem, 2017; Zenke et al., 2017; Nguyen et al., 2017; Titsias et al., 2020; Kolouri et al., 2020; Kao et al., 2021).”
A Unified and General Framework for Continual Learning (2024, methodology): “Natural Gradient CL as A Special Case Natural Gradient CL (Osawa et al., 2019; Kao et al., 2021) (NCL) is a Bayesian-based CL method.”
How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Composition (2022, methodology): “, 2018) with a novel projection method onto the trust region over the posterior of previous tasks (Kao et al., 2021).”