Generalized Class Incremental Learning

By Fei Mi, Lingjing Kong, Tao Lin, Kaicheng Yu, Boi Faltings. 2020

48 citations recorded by the source corpus.

Many real-world machine learning systems require the ability to continually learn new knowledge. Class incremental learning receives increasing attention recently as a solution towards this goal. However, existing methods often introduce some assumptions to simplify the problem setting, which rarely holds in real-world scenarios. In this paper, we formulate a Generalized Class Incremental Learning (GCIL) framework to systematically alleviate these restrictions, and introduce several novel realistic incremental learning scenarios. In addition, we propose a simple yet effective method, namely ReMix, which combines Exemplar Replay (ER) and Mixup to deal with different challenges in realistic GCIL setups. We demonstrate on CIFAR-100 that ReMix outperforms the state-of-the-art methods in different GCIL setups by significant margins without introducing additional computation cost.

Subfields: Artificial Intelligence, Cancer Research, Computer Vision and Pattern Recognition, Meta & Continual Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Hang Su
2. Jun Zhu
3. Gyuhak Kim
4. Xingxing Zhang
5. Liyuan Wang
6. Bahram Zonooz
7. Elahe Arani
8. Changnan Xiao

In Cancer Research

1. Jun Zhu

In Computer Vision and Pattern Recognition

1. Hang Su
2. Gyuhak Kim
3. Jun Zhu
4. Liyuan Wang
5. Xingxing Zhang
6. Changnan Xiao
7. Sepideh Esmaeilpour
8. Boi Faltings

In Meta & Continual Learning

1. Xingxing Zhang
2. Liyuan Wang
3. Hang Su
4. Jun Zhu
5. Gyuhak Kim
6. Changnan Xiao
7. Bahram Zonooz
8. Elahe Arani

Citations in the corpus

• A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, background): “Some recent work has focused on combinatorial scenarios [18], [40], [41], [42], [43] in order to simulate real-world complexity.”
• Learning Fast, Learning Slow: A General Continual Learning Method based on Complementary Learning System (2022, methodology): “The Generalized Class Incremental Learning (GCIL; Mi et al. (2020)) is the closest to the real-world scenario as it utilizes probabilistic modeling to sample the classes and data distributions in each task.”
• Continual Learning Based on OOD Detection and Task Masking (2022, background): “, a network of experts (Aljundi, Chakravarty, and Tuytelaars 2017) and generalized CL (Mi et al. 2020), etc.”
• ADER: Adaptively Distilled Exemplar Replay Towards Continual Learning for Session-based Recommendation (2020, methodology): “Besides selecting exemplars uniformly, [31] incorporates the Herding technique [38] to select exemplars, and it soon becomes popular [3, 14, 28, 39, 41].”
• Retrospective Adversarial Replay for Continual Learning (2022, methodology): “The proposed MixUp strategy is different from [21, 40] as we apply mixup among the replay samples and then generate RAR perturbed samples anchored around them.”