Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
102 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

CORE: Mitigating Catastrophic Forgetting in Continual Learning through Cognitive Replay (2402.01348v2)

Published 2 Feb 2024 in cs.LG and cs.AI

Abstract: This paper introduces a novel perspective to significantly mitigate catastrophic forgetting in continuous learning (CL), which emphasizes models' capacity to preserve existing knowledge and assimilate new information. Current replay-based methods treat every task and data sample equally and thus can not fully exploit the potential of the replay buffer. In response, we propose COgnitive REplay (CORE), which draws inspiration from human cognitive review processes. CORE includes two key strategies: Adaptive Quantity Allocation and Quality-Focused Data Selection. The former adaptively modulates the replay buffer allocation for each task based on its forgetting rate, while the latter guarantees the inclusion of representative data that best encapsulates the characteristics of each task within the buffer. Our approach achieves an average accuracy of 37.95% on split-CIFAR10, surpassing the best baseline method by 6.52%. Additionally, it significantly enhances the accuracy of the poorest-performing task by 6.30% compared to the top baseline. Code is available at https://github.com/sterzhang/CORE.

PDF HTML Abstract
Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (34)
  1. (2012, Sep). Using spacing to enhance diverse forms of learning: Review of recent research and implications for instruction. Educational Psychology Review, 369–378. doi: 10.1007/s10648-012-9205-z
  2. (2020, Jan). Lifelong machine learning. In Transfer learning (p. 196–208). doi: 10.1017/9781139061773.016
  3. (1972, Dec). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 671–684. doi: 10.1016/s0022-5371(72)80001-x
  4. (2021). A continual learning survey: Defying forgetting in classification tasks. IEEE transactions on pattern analysis and machine intelligence, 44(7), 3366–3385.
  5. (2022). A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7), 3366-3385. doi: 10.1109/TPAMI.2021.3057446
  6. (2022). Orthogonal representations for robust context-dependent task performance in brains and neural networks. Neuron, 110(7), 1258–1270.
  7. French, R.  (1999, Apr). Catastrophic forgetting in connectionist networks. Trends in Cognitive Sciences, 128–135. doi: 10.1016/s1364-6613(99)01294-2
  8. French, R. M.  (1999). Catastrophic forgetting in connectionist networks. Trends in cognitive sciences, 3(4), 128–135.
  9. Huszár, F.  (2018). Note on the quadratic penalties in elastic weight consolidation. Proceedings of the National Academy of Sciences, 115(11), E2496–E2497.
  10. (2009). Learning multiple layers of features from tiny images.
  11. (2017). Learning without forgetting. IEEE transactions on pattern analysis and machine intelligence, 40(12), 2935–2947.
  12. (1990). Levels of processing: A retrospective commentary on a framework for memory research. Canadian Journal of Psychology/Revue canadienne de psychologie, 44(1), 87.
  13. (2018). Alleviating catastrophic forgetting using context-dependent gating and synaptic stabilization. Proceedings of the National Academy of Sciences, 115(44), E10467–E10475.
  14. (2023). Multitask learning via interleaving: A neural network investigation. In Proceedings of the annual meeting of the cognitive science society (Vol. 45).
  15. (1989). Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation (Vol. 24, pp. 109–165). Elsevier.
  16. (1943). A logical calculus of the ideas immanent in nervous activity. The bulletin of mathematical biophysics, 5, 115–133.
  17. (2013, Dec). Playing atari with deep reinforcement learning. arXiv: Learning,arXiv: Learning.
  18. (2015). Human-level control through deep reinforcement learning. nature, 518(7540), 529–533.
  19. (2017). icarl: Incremental classifier and representation learning. In Proceedings of the ieee conference on computer vision and pattern recognition (pp. 2001–2010).
  20. (2019). Experience replay for continual learning. Advances in Neural Information Processing Systems, 32.
  21. (2016). Progressive neural networks. arXiv preprint arXiv:1606.04671.
  22. (2016). Forgetting patterns differentiate between two forms of memory representation. Psychological science, 27(6), 810–820.
  23. (2006). Shallow processing and underspecification. Discourse Processes, 42(2), 99–108.
  24. (2023). Brain-inspired learning in artificial neural networks: a review. arXiv preprint arXiv:2305.11252.
  25. (2017). Continual learning with deep generative replay. Advances in neural information processing systems, 30.
  26. (2017). Efficient processing of deep neural networks: A tutorial and survey. Proceedings of the IEEE, 105(12), 2295–2329.
  27. (2019, 09–15 Jun). EfficientNet: Rethinking model scaling for convolutional neural networks. In K. Chaudhuri  R. Salakhutdinov (Eds.), Proceedings of the 36th international conference on machine learning (Vol. 97, pp. 6105–6114). PMLR. Retrieved from https://proceedings.mlr.press/v97/tan19a.html
  28. (2022). Gcr: Gradient coreset based replay buffer selection for continual learning. In Proceedings of the ieee/cvf conference on computer vision and pattern recognition (pp. 99–108).
  29. (2019). Three scenarios for continual learning. arXiv preprint arXiv:1904.07734.
  30. (2022). Three types of incremental learning. Nature Machine Intelligence, 4, 1185–1197.
  31. (2023). A comprehensive survey of continual learning: Theory, method and application. arXiv preprint arXiv:2302.00487.
  32. (2015, Jun). Understanding neural networks through deep visualization. arXiv: Computer Vision and Pattern Recognition,arXiv: Computer Vision and Pattern Recognition.
  33. (2014). Visualizing and understanding convolutional networks. In Computer vision–eccv 2014: 13th european conference, zurich, switzerland, september 6-12, 2014, proceedings, part i 13 (pp. 818–833).
  34. (2017). Continual learning through synaptic intelligence. In International conference on machine learning (pp. 3987–3995).
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (5)
  1. Jianshu Zhang (36 papers)
  2. Yankai Fu (5 papers)
  3. Ziheng Peng (1 paper)
  4. Dongyu Yao (8 papers)
  5. Kun He (177 papers)
Github Logo Streamline Icon: https://streamlinehq.com

GitHub

  1. GitHub - sterzhang/CORE: Official Repo for CORE (CogSci 2024 Oral) (4 stars)
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets