Generative Models from the perspective of Continual Learning
This work addresses the problem of selecting effective generative models for continual learning in image generation, though it is incremental as it compares existing methods.
The paper evaluated generative models for continual learning on sequential image tasks, finding that original GANs performed best and generative replay was the top strategy, with satisfactory results on MNIST and Fashion MNIST but instability on CIFAR10.
Which generative model is the most suitable for Continual Learning? This paper aims at evaluating and comparing generative models on disjoint sequential image generation tasks. We investigate how several models learn and forget, considering various strategies: rehearsal, regularization, generative replay and fine-tuning. We used two quantitative metrics to estimate the generation quality and memory ability. We experiment with sequential tasks on three commonly used benchmarks for Continual Learning (MNIST, Fashion MNIST and CIFAR10). We found that among all models, the original GAN performs best and among Continual Learning strategies, generative replay outperforms all other methods. Even if we found satisfactory combinations on MNIST and Fashion MNIST, training generative models sequentially on CIFAR10 is particularly instable, and remains a challenge. Our code is available online \footnote{\url{https://github.com/TLESORT/Generative\_Continual\_Learning}}.