Balanced softmax cross-entropy for incremental learning with and without memory

TL;DR

This paper introduces a balanced softmax cross-entropy method to improve incremental learning in neural networks, effectively reducing catastrophic forgetting with or without memory, and demonstrates its advantages on multiple benchmarks.

Contribution

It proposes a novel balanced softmax cross-entropy approach that enhances incremental learning performance and can be combined with existing methods or used independently without memory.

Findings

Improves accuracy in class-incremental learning tasks.

Reduces catastrophic forgetting effectively.

Achieves competitive results without memory.

Abstract

When incrementally trained on new classes, deep neural networks are subject to catastrophic forgetting which leads to an extreme deterioration of their performance on the old classes while learning the new ones. Using a small memory containing few samples from past classes has shown to be an effective method to mitigate catastrophic forgetting. However, due to the limited size of the replay memory, there is a large imbalance between the number of samples for the new and the old classes in the training dataset resulting in bias in the final model. To address this issue, we propose to use the Balanced Softmax Cross-Entropy and show that it can be seamlessly combined with state-of-the-art approaches for class-incremental learning in order to improve their accuracy while also potentially decreasing the computational cost of the training procedure. We further extend this approach to the more demanding class-incremental learning without memory setting and achieve competitive results with memory-based approaches. Experiments on the challenging ImageNet, ImageNet-Subset and CIFAR100 benchmarks with various settings demonstrate the benefits of our approach.