Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
162 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Online Hyperparameter Optimization for Class-Incremental Learning (2301.05032v2)

Published 11 Jan 2023 in cs.LG and cs.CV

Abstract: Class-incremental learning (CIL) aims to train a classification model while the number of classes increases phase-by-phase. An inherent challenge of CIL is the stability-plasticity tradeoff, i.e., CIL models should keep stable to retain old knowledge and keep plastic to absorb new knowledge. However, none of the existing CIL models can achieve the optimal tradeoff in different data-receiving settings--where typically the training-from-half (TFH) setting needs more stability, but the training-from-scratch (TFS) needs more plasticity. To this end, we design an online learning method that can adaptively optimize the tradeoff without knowing the setting as a priori. Specifically, we first introduce the key hyperparameters that influence the trade-off, e.g., knowledge distillation (KD) loss weights, learning rates, and classifier types. Then, we formulate the hyperparameter optimization process as an online Markov Decision Process (MDP) problem and propose a specific algorithm to solve it. We apply local estimated rewards and a classic bandit algorithm Exp3 to address the issues when applying online MDP methods to the CIL protocol. Our method consistently improves top-performing CIL methods in both TFH and TFS settings, e.g., boosting the average accuracy of TFH and TFS by 2.2 percentage points on ImageNet-Full, compared to the state-of-the-art.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (41)
  1. Conditional channel gated networks for task-aware continual learning. In CVPR, pages 3931–3940, 2020.
  2. Analysis of thompson sampling for the multi-armed bandit problem. In COLT, volume 23, pages 39.1–39.26, 2012.
  3. Terry Anderson. The theory and practice of online learning. Athabasca University Press, 2008.
  4. The nonstochastic multiarmed bandit problem. SIAM journal on computing, 32(1):48–77, 2002.
  5. Ucb revisited: Improved regret bounds for the stochastic multi-armed bandit problem. Periodica Mathematica Hungarica, 61(1-2):55–65, 2010.
  6. Automatic differentiation in machine learning: a survey. JMLR, 18:1–43, 2018.
  7. Il2m: Class incremental learning with dual memory. In CVPR, pages 583–592, 2019.
  8. Podnet: Pooled outputs distillation for small-tasks incremental learning. In ECCV, pages 86–102, 2020.
  9. Experts in a markov decision process. In NIPS, pages 401–408, 2005.
  10. Online markov decision processes. Mathematics of Operations Research, 34(3):726–736, 2009.
  11. Bilevel programming for hyperparameter optimization and meta-learning. In ICML, pages 1568–1577, 2018.
  12. Deep residual learning for image recognition. In CVPR, pages 770–778, 2016.
  13. Distilling the knowledge in a neural network. arXiv, 1503.02531, 2015.
  14. Learning a unified classifier incrementally via rebalancing. In CVPR, pages 831–839, 2019.
  15. Distilling causal effect of data in class-incremental learning. In CVPR, pages 3957–3966, 2021.
  16. Energy-based latent aligner for incremental learning. In CVPR, pages 7452–7461, 2022.
  17. Learning multiple layers of features from tiny images. Technical report, University of Toronto, 2009.
  18. Online optimal control with affine constraints. In AAAI, pages 8527–8537, 2021.
  19. Online learning for markov decision processes in nonstationary environments: A dynamic regret analysis. In 2019 American Control Conference (ACC), pages 1232–1237. IEEE, 2019.
  20. Online markov decision processes with time-varying transition probabilities and rewards. In ICML workshop on Real-world Sequential Decision Making, 2019.
  21. Learning without forgetting. In ECCV, pages 614–629, 2016.
  22. Adaptive aggregation networks for class-incremental learning. In CVPR, pages 2544–2553, 2021.
  23. Rmm: Reinforced memory management for class-incremental learning. In NeurIPS, pages 3478–3490, 2021.
  24. Mnemonics training: Multi-class incremental learning without forgetting. In CVPR, pages 12245–12254, 2020.
  25. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of Learning and Motivation, volume 24, pages 109–165. Elsevier, 1989.
  26. Catastrophic interference is eliminated in pre-trained networks. In CogSci, 1993.
  27. Looking back on learned experiences for class/task incremental learning. In ICLR, 2022.
  28. Gdumb: A simple approach that questions our progress in continual learning. In ECCV, pages 524–540, 2020.
  29. R. Ratcliff. Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions. Psychological Review, 97:285–308, 1990.
  30. iCaRL: Incremental classifier and representation learning. In CVPR, pages 5533–5542, 2017.
  31. Imagenet large scale visual recognition challenge. IJCV, 115(3):211–252, 2015.
  32. Progressive neural networks. arXiv, 1606.04671, 2016.
  33. Grad-cam: Visual explanations from deep networks via gradient-based localization. In CVPR, pages 618–626, 2017.
  34. Continual learning with deep generative replay. In NIPS, pages 2990–2999, 2017.
  35. Prototypical networks for few-shot learning. In NIPS, pages 4077–4087, 2017.
  36. Topology-preserving class-incremental learning. In ECCV, pages 254–270, 2020.
  37. Encoder based lifelong learning. In ICCV, pages 1329–1337, 2017.
  38. Foster: Feature boosting and compression for class-incremental learning. In ECCV, pages 398–414, 2022.
  39. Reinforced continual learning. In NeurIPS, pages 899–908, 2018.
  40. Der: Dynamically expandable representation for class incremental learning. In CVPR, pages 3014–3023, 2021.
  41. Semantic drift compensation for class-incremental learning. In CVPR, pages 6982–6991, 2020.
Citations (29)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now