CLA: Latent Alignment for Online Continual Self-Supervised Learning

TL;DR

This paper introduces Continual Latent Alignment (CLA), a novel SSL method for online continual learning that aligns current and past representations to reduce forgetting and improve training efficiency and final performance.

Contribution

CLA is a new SSL strategy for online continual learning that mitigates forgetting by aligning current and past representations, enhancing convergence and final accuracy.

Findings

CLA speeds up training convergence in online scenarios.

CLA outperforms state-of-the-art methods under the same computational budget.

Using CLA as pretraining improves final performance.

Abstract

Self-supervised learning (SSL) is able to build latent representations that generalize well to unseen data. However, only a few SSL techniques exist for the online CL setting, where data arrives in small minibatches, the model must comply with a fixed computational budget, and task boundaries are absent. We introduce Continual Latent Alignment (CLA), a novel SSL strategy for Online CL that aligns the representations learned by the current model with past representations to mitigate forgetting. We found that our CLA is able to speed up the convergence of the training process in the online scenario, outperforming state-of-the-art approaches under the same computational budget. Surprisingly, we also discovered that using CLA as a pretraining protocol in the early stages of pretraining leads to a better final performance when compared to a full i.i.d. pretraining.