Self-Supervised Pseudodata Filtering for Improved Replay with Sub-Optimal Generators
Keywords: continual learning, catastrophic forgetting, generative replay, feature replay, deep learning
TL;DR: The paper shows how using the classifier's confidence (softmax value) to select good-quality generated samples improves the performance of generative replay methods operating either on full images or on features.
Abstract: Continual learning of a sequence of tasks without forgetting previously acquired knowledge is one of the main challenges faced by modern deep neural networks. In the class-incremental scenario (aka open-set learning), one of the most difficult continual learning problems, new classes are presented to a classifier over time. The model needs to be able to learn and recognize these new classes while also retaining its knowledge of previously witnessed ones. A common approach is to make it revisit the old classes or their features in some form, either by analysing stored exemplars or by using artificially generated samples. The latter approach, Generative Replay, usually relies on a separate generator trained alongside the main classifier. Since the generator also needs to learn continually, it is usually retrained on every task, using its own generated samples as training data representing older classes. This can lead to error propagation and accumulating features unimportant or confusing for the classifier, reducing the overall performance for larger numbers of tasks. We propose a simple filtering mechanism for mitigating this issue – whenever pseudodata is generated for a new task, the classifier can reject samples it is not able to classify with sufficient confidence, thus preventing both models from retraining on poor-quality data. We tested the filter on several datasets, including real-life images, using various combinations of models, as the method can be applied regardless of the network architectures. We show that filtering improves the classifier's accuracy and provide statistical analysis of the results.
Primary Area: transfer learning, meta learning, and lifelong learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10882
Loading