Bypassing Logits Bias in Online Class-Incremental Learning with a Generative Framework

TL;DR

This paper introduces a generative framework for online class-incremental learning that bypasses the logits bias problem of softmax classifiers, leading to improved performance and reduced catastrophic forgetting.

Contribution

It proposes a novel generative feature-space-based approach that replaces softmax classifiers and combines generative and discriminative losses for better online continual learning.

Findings

Outperforms state-of-the-art replay-based methods on multiple benchmarks.

Effectively reduces catastrophic forgetting in online learning scenarios.

Works well on newly introduced task-free datasets.

Abstract

Continual learning requires the model to maintain the learned knowledge while learning from a non-i.i.d data stream continually. Due to the single-pass training setting, online continual learning is very challenging, but it is closer to the real-world scenarios where quick adaptation to new data is appealing. In this paper, we focus on online class-incremental learning setting in which new classes emerge over time. Almost all existing methods are replay-based with a softmax classifier. However, the inherent logits bias problem in the softmax classifier is a main cause of catastrophic forgetting while existing solutions are not applicable for online settings. To bypass this problem, we abandon the softmax classifier and propose a novel generative framework based on the feature space. In our framework, a generative classifier which utilizes replay memory is used for inference, and the training objective is a pair-based metric learning loss which is proven theoretically to optimize the feature space in a generative way. In order to improve the ability to learn new data, we further propose a hybrid of generative and discriminative loss to train the model. Extensive experiments on several benchmarks, including newly introduced task-free datasets, show that our method beats a series of state-of-the-art replay-based methods with discriminative classifiers, and reduces catastrophic forgetting consistently with a remarkable margin.