A comprehensive study of class incremental learning algorithms for visual tasks. (March 2021)

• Record Type: Journal Article
• Authors: Belouadah, Eden; Popescu, Adrian; Kanellos, Ioannis
• Is Part Of: Neural networks. Volume 135(2021)
• Page Start: 38
• Page End: 54
• Publication Date: 2021-03
• Subjects: Incremental learning -- Catastrophic forgetting -- Imbalanced learning -- Image classification -- Convolutional neural networks; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2020.12.003
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: The ability of artificial agents to increment their capabilities when confronted with new data is an open challenge in artificial intelligence. The main challenge faced in such cases is catastrophic forgetting, i.e., the tendency of neural networks to underfit past data when new ones are ingested. A first group of approaches tackles forgetting by increasing deep model capacity to accommodate new knowledge. A second type of approaches fix the deep model size and introduce a mechanism whose objective is to ensure a good compromise between stability and plasticity of the model. While the first type of algorithms were compared thoroughly, this is not the case for methods which exploit a fixed size model. Here, we focus on the latter, place them in a common conceptual and experimental framework and propose the following contributions: (1) define six desirable properties of incremental learning algorithms and analyze them according to these properties, (2) introduce a unified formalization of the class-incremental learning problem, (3) propose a common evaluation framework which is more thorough than existing ones in terms of number of datasets, size of datasets, size of bounded memory and number of incremental states, (4) investigate the usefulness of herding for past exemplars selection, (5) provide experimental evidence that it is possible to obtain competitive performance without the use of knowledge distillation to tackle catastrophic forgetting and (6) facilitate reproducibility by integrating all tested methods in a common open-source repository. The main experimental finding is that none of the existing algorithms achieves the best results in all evaluated settings. Important differences arise notably if a bounded memory of past classes is allowed or not. Highlights: Incremental learning algorithms are improved by casting the problem as an imbalanced learning case. Competitive performance can be achieved without the widely used knowledge distillation component. Herding-based exemplar selection for past classes clearly outperforms random selection. Fine-tuning based methods are better when a memory of the past is allowed. Fixed-representation based methods are better without a memory of the past.