Generative negative replay for continual learning.

Autor: Graffieti G; Department of Computer Science and Engineering, University of Bologna, Italy. Electronic address: gabriele.graffieti@unibo.it., Maltoni D; Department of Computer Science and Engineering, University of Bologna, Italy. Electronic address: davide.maltoni@unibo.it., Pellegrini L; Department of Computer Science and Engineering, University of Bologna, Italy. Electronic address: l.pellegrini@unibo.it., Lomonaco V; Department of Computer Science, University of Pisa, Italy. Electronic address: vincenzo.lomonaco@unipi.it.
Jazyk: angličtina
Zdroj: Neural networks : the official journal of the International Neural Network Society [Neural Netw] 2023 May; Vol. 162, pp. 369-383. Date of Electronic Publication: 2023 Mar 09.
DOI: 10.1016/j.neunet.2023.03.006
Abstrakt: Learning continually is a key aspect of intelligence and a necessary ability to solve many real-life problems. One of the most effective strategies to control catastrophic forgetting, the Achilles' heel of continual learning, is storing part of the old data and replaying them interleaved with new experiences (also known as the replay approach). Generative replay, which is using generative models to provide replay patterns on demand, is particularly intriguing, however, it was shown to be effective mainly under simplified assumptions, such as simple scenarios and low-dimensional data. In this paper, we show that, while the generated data are usually not able to improve the classification accuracy for the old classes, they can be effective as negative examples (or antagonists) to better learn the new classes, especially when the learning experiences are small and contain examples of just one or few classes. The proposed approach is validated on complex class-incremental and data-incremental continual learning scenarios (CORe50 and ImageNet-1000) composed of high-dimensional data and a large number of training experiences: a setup where existing generative replay approaches usually fail.
Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
(Copyright © 2023 Elsevier Ltd. All rights reserved.)
Databáze: MEDLINE
Externí odkaz: