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

Learnability and Algorithm for Continual Learning (2306.12646v1)

Published 22 Jun 2023 in cs.LG and cs.CV

Abstract: This paper studies the challenging continual learning (CL) setting of Class Incremental Learning (CIL). CIL learns a sequence of tasks consisting of disjoint sets of concepts or classes. At any time, a single model is built that can be applied to predict/classify test instances of any classes learned thus far without providing any task related information for each test instance. Although many techniques have been proposed for CIL, they are mostly empirical. It has been shown recently that a strong CIL system needs a strong within-task prediction (WP) and a strong out-of-distribution (OOD) detection for each task. However, it is still not known whether CIL is actually learnable. This paper shows that CIL is learnable. Based on the theory, a new CIL algorithm is also proposed. Experimental results demonstrate its effectiveness.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (76)
  1. Conditional channel gated networks for task-aware continual learning. In CVPR, pp.  3931–3940, 2020.
  2. Uncertainty-based continual learning with adaptive regularization. In NeurIPS, 2019.
  3. Expert gate: Lifelong learning with a network of experts. In CVPR, 2017.
  4. Online continual learning with maximal interfered retrieval. In NeurIPS, 2019.
  5. Generalisation guarantees for continual learning with orthogonal gradient descent. Lifelong Learning Workshop at the ICML, 2020.
  6. Dark experience for general continual learning: a strong, simple baseline. In NeurIPS, 2020.
  7. End-to-end incremental learning. In ECCV, pp.  233–248, 2018.
  8. Co2l: Contrastive continual learning. In ICCV, 2021.
  9. Efficient lifelong learning with a-gem. In ICLR, 2019a.
  10. Continual learning with tiny episodic memories. 2019b.
  11. Continual learning in low-rank orthogonal subspaces, 2020.
  12. Using hindsight to anchor past knowledge in continual learning. Proceedings of the AAAI Conference on Artificial Intelligence, 35(8):6993–7001, May 2021. URL https://ojs.aaai.org/index.php/AAAI/article/view/16861.
  13. Lifelong machine learning. Synthesis Lectures on Artificial Intelligence and Machine Learning, 12(3):1–207, 2018.
  14. Learning without memorizing. In CVPR, 2019.
  15. Is out-of-distribution detection learnable? aNeurIPS-2022, 2022.
  16. Online continual learning through mutual information maximization. In International Conference on Machine Learning, pp. 8109–8126. PMLR, 2022.
  17. Lifelong machine learning with deep streaming linear discriminant analysis. In CVPR Workshop on Continual Learning, 2020.
  18. A baseline for detecting misclassified and out-of-distribution examples in neural networks. arXiv preprint arXiv:1610.02136, 2016.
  19. Posterior meta-replay for continual learning. NeurIPS, 34, 2021.
  20. Learning a unified classifier incrementally via rebalancing. In CVPR, pp.  831–839, 2019.
  21. Parameter-efficient transfer learning for nlp. In International Conference on Machine Learning, pp. 2790–2799. PMLR, 2019.
  22. Overcoming catastrophic forgetting for continual learning via model adaptation. In ICLR, 2019.
  23. Compacting, picking and growing for unforgetting continual learning. In NeurIPS, volume 32, 2019.
  24. Deep Generative Dual Memory Network for Continual Learning. arXiv preprint arXiv:1710.10368, 2017.
  25. Continual learning of a mixed sequence of similar and dissimilar tasks. In NeurIPS, 2020.
  26. Achieving forgetting prevention and knowledge transfer in continual learning. NeurIPS, 2021.
  27. FearNet: Brain-Inspired Model for Incremental Learning. In ICLR, 2018.
  28. Continual learning via principal components projection, 2020. URL https://openreview.net/forum?id=SkxlElBYDS.
  29. A multi-head model for continual learning via out-of-distribution replay. In Conference on Lifelong Learning Agents, pp.  548–563. PMLR, 2022a.
  30. A theoretical study on solving continual learning. NeurIPS-2022, 2022b.
  31. Open-world continual learning: Unifying novelty detection and continual learning. arXiv:2304.10038 [cs.LG], 2023.
  32. Overcoming catastrophic forgetting in neural networks. Proceedings of the national academy of sciences, 114(13):3521–3526, 2017.
  33. Learning multiple layers of features from tiny images. Technical Report TR-2009, University of Toronto, Toronto., 2009.
  34. Tiny imagenet visual recognition challenge, 2015.
  35. Overcoming catastrophic forgetting with unlabeled data in the wild. In CVPR, 2019.
  36. Continual learning in the teacher-student setup: Impact of task similarity. In International Conference on Machine Learning, pp. 6109–6119. PMLR, 2021.
  37. Learning Without Forgetting. In ECCV, pp.  614–629. Springer, 2016.
  38. Beyond not-forgetting: Continual learning with backward knowledge transfer. NeurIPS-2022, 2022.
  39. Ai autonomy: Self-initiated open-world continual learning and adaptation. AI Magazine, 2023.
  40. More classifiers, less forgetting: A generic multi-classifier paradigm for incremental learning. In ECCV, pp.  699–716. Springer International Publishing, 2020a. doi: 10.1007/978-3-030-58574-7˙42. URL https://doi.org/10.1007/978-3-030-58574-7_42.
  41. Mnemonics training: Multi-class incremental learning without forgetting. In CVPR, 2020b.
  42. Adaptive aggregation networks for class-incremental learning. In CVPR, 2021.
  43. Gradient Episodic Memory for Continual Learning. In NeurIPS, pp.  6470–6479, 2017.
  44. PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning. arXiv preprint arXiv:1711.05769, 2017.
  45. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, pp. 109–165. Elsevier, 1989.
  46. Learning to remember: A synaptic plasticity driven framework for continual learning. In CVPR, pp.  11321–11329, 2019.
  47. Continual learning via local module composition. Advances in Neural Information Processing Systems, 34:30298–30312, 2021.
  48. Continual learning with foundation models: An empirical study of latent replay. Conference on Lifelong Learning Agents, 2022.
  49. A pac-bayesian bound for lifelong learning. In International Conference on Machine Learning, pp. 991–999. PMLR, 2014.
  50. Learning transferable visual models from natural language supervision. arXiv preprint arXiv:2103.00020, 2021.
  51. An adaptive random path selection approach for incremental learning, 2020a.
  52. itaml: An incremental task-agnostic meta-learning approach. In CVPR, 2020b.
  53. iCaRL: Incremental classifier and representation learning. In CVPR, pp.  5533–5542, 2017.
  54. Online structured laplace approximations for overcoming catastrophic forgetting. In NeurIPS, 2018.
  55. Experience replay for continual learning. In NeurIPS, 2019.
  56. Complementary learning for overcoming catastrophic forgetting using experience replay. In IJCAI, 2019.
  57. Progressive neural networks. arXiv preprint arXiv:1606.04671, 2016.
  58. Progress & compress: A scalable framework for continual learning. arXiv preprint arXiv:1805.06370, 2018.
  59. Continual learning in generative adversarial nets. arXiv preprint arXiv:1705.08395, 2017.
  60. Overcoming catastrophic forgetting with hard attention to the task. In ICML, 2018.
  61. Continual learning with deep generative replay. In NIPS, pp.  2994–3003, 2017.
  62. Training data-efficient image transformers & distillation through attention. In International Conference on Machine Learning, pp. 10347–10357. PMLR, 2021.
  63. Three scenarios for continual learning. arXiv preprint arXiv:1904.07734, 2019.
  64. Continual learning with hypernetworks. ICLR, 2020.
  65. Vim: Out-of-distribution with virtual-logit matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.  4921–4930, 2022a.
  66. Memory replay with data compression for continual learning. Proceedings of International Conference on Learning Representations (ICLR), 2022b.
  67. Learning to prompt for continual learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.  139–149, 2022c.
  68. Supermasks in superposition. In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M. F., and Lin, H. (eds.), NeurIPS, 2020.
  69. Memory replay gans: Learning to generate new categories without forgetting. In NeurIPS, 2018.
  70. Class-incremental learning with strong pre-trained models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  9601–9610, June 2022.
  71. Large scale incremental learning. In CVPR, 2019.
  72. Reinforced continual learning. In NeurIPS, 2018.
  73. Der: Dynamically expandable representation for class incremental learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.  3014–3023, 2021.
  74. Continuous learning of context-dependent processing in neural networks. Nature Machine Intelligence, 2019.
  75. Continual learning through synaptic intelligence. In ICML, pp.  3987–3995, 2017.
  76. Prototype augmentation and self-supervision for incremental learning. In CVPR, 2021.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Gyuhak Kim (9 papers)
  2. Changnan Xiao (12 papers)
  3. Tatsuya Konishi (7 papers)
  4. Bing Liu (212 papers)
Citations (14)
View on Semantic Scholar

Summary

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

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