Abstract:Regularization-based methods are beneficial to alleviate the catastrophic forgetting problem in class-incremental learning. With the absence of old task images, they often assume that old knowledge is well preserved if the classifier produces similar output on new images. In this paper, we find that their effectiveness largely depends on the nature of old classes: they work well on classes that are easily distinguishable between each other but may fail on more fine-grained ones, e.g., boy and girl. In spirit, such methods project new data onto the feature space spanned by the weight vectors in the fully connected layer, corresponding to old classes. The resulting projections would be similar on fine-grained old classes, and as a consequence the new classifier will gradually lose the discriminative ability on these classes. To address this issue, we propose a memory-free generative replay strategy to preserve the fine-grained old classes characteristics by generating representative old images directly from the old classifier and combined with new data for new classifier training. To solve the homogenization problem of the generated samples, we also propose a diversity loss that maximizes Kullback Leibler (KL) divergence between generated samples. Our method is best complemented by prior regularization-based methods proved to be effective for easily distinguishable old classes. We validate the above design and insights on CUB-200-2011, Caltech-101, CIFAR-100 and Tiny ImageNet and show that our strategy outperforms existing memory-free methods with a clear margin. Code is available at <a class="link-external link-https" href="https://github.com/xmengxin/MFGR" rel="external noopener nofollow">this https URL</a>

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

Boosting Separated Softmax with Discrimination for Class Incremental Learning

Learning a Unified Classifier Incrementally via Rebalancing

Memory-Free Generative Replay For Class-Incremental Learning

Discrimination Correction and Balance for Class-Incremental Learning

Large Scale Incremental Learning

Rethinking Class-Incremental Learning from a Dynamic Imbalanced Learning Perspective

Mnemonics Training: Multi-Class Incremental Learning Without Forgetting

Class-Incremental Learning Based on Big Dataset Pre-Trained Models

Incremental Learning Using Conditional Adversarial Networks

Continual Recognition with Adaptive Memory Update.

An Efficient Replay for Class-Incremental Learning with Pre-trained Models

Incremental Learning of Structured Memory via Closed-Loop Transcription

Feature Expansion and enhanced Compression for Class Incremental Learning

An Efficient Strategy for Catastrophic Forgetting Reduction in Incremental Learning

Incremental Learning Via Rate Reduction

Class-incremental learning with Balanced Embedding Discrimination Maximization

Adaptive knowledge transfer for class incremental learning

More Task-Balanced Class-Incremental Learning

Online Class-incremental Continual Learning with Maximum Entropy Memory Update