Incremental Concept Learning via Online Generative Memory Recall

Abstract

The ability to learn more and more concepts over time from incrementally arriving data is essential for the development of a life-long learning system. However, deep neural networks often suffer from forgetting previously learned concepts when continually learning new concepts, which is known as catastrophic forgetting problem. The main reason for catastrophic forgetting is that the past concept data is not available and neural weights are changed during incrementally learning new concepts. In this paper, we propose a pseudo-rehearsal based class incremental learning approach to make neural networks capable of continually learning new concepts. We use a conditional generative adversarial network to consolidate old concepts memory and recall pseudo samples during learning new concepts and a balanced online memory recall strategy is to maximally maintain old memories. And we design a comprehensible incremental concept learning network as well as a concept contrastive loss to alleviate the magnitude of neural weights change. We evaluate the proposed approach on MNIST, Fashion-MNIST and SVHN datasets and compare with other rehearsal based approaches. The extensive experiments demonstrate the effectiveness of our approach.

SAM-kNN Regressor for Online Learning in Water Distribution Networks

“…in the form of a potentially infinite streams of data. Incremental learning has made great strides in recent years [5], with most applications set in a classification environment [18,10,12,9,7,21].…”

FFNB: Forgetting-Free Neural Blocks for Deep Continual Visual Learning

“…Pseudo-replay models [23] including deep generative networks [12], [22], [25], [48], [56] have also been investigated in the literature as alternative solutions that prevent storing exemplars. The particular method in [31] combines an explicit and an implicit memory where the former captures features from observed tasks and the latter corresponds to a discriminator and a generator similarly to deep generative replay. Other continual learning approaches employ coresets [42] to characterize the key information from different tasks.…”