Balanced softmax cross-entropy for incremental learning with and without memory
This work addresses incremental learning for AI systems that need to adapt to new classes over time, offering an incremental improvement by mitigating bias from memory constraints.
The paper tackles catastrophic forgetting in class-incremental learning by proposing Balanced Softmax Cross-Entropy to address class imbalance from limited replay memory, improving accuracy and reducing computational cost in memory-based and memory-free settings.
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.