Research map / Continual learning / Class-incremental
Online Class-Incremental Continual Learning with Adversarial Shapley Value
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace random retrieval and reservoir updates with ASER’s combined Shapley-based retrieval and update | ER · Split Mini-ImageNet, M=1k | average accuracy: 8.7\pm0.4 -> 11.7\pm0.8 | Helped |
| Replace random retrieval and reservoir updates with ASER’s combined Shapley-based retrieval and update | ER · Split CIFAR-100, M=1k | average accuracy: 11.2\pm0.4 -> 12.3\pm0.4 | Helped |
| Replace random retrieval and reservoir updates with ASER’s combined Shapley-based retrieval and update | ER · Split CIFAR-10, M=0.2k | average accuracy: 26.4\pm1.0 -> 27.8\pm1.0 | Mixed |
| Use mean ASV rather than max-minus-min ASV scoring | ASER · Split Mini-ImageNet and Split CIFAR-100, M=1k | average accuracy: 11.7\pm0.7 -> 12.2\pm0.8; 12.3\pm0.4 -> 14.0\pm0.4 | Helped |
| Use KNN-SV memory updates rather than reservoir updates (ablation) | ER with random retrieval · Split Mini-ImageNet and Split CIFAR-100, M=1k | average accuracy: 8.7\pm0.4 -> 13.4\pm0.8; 11.2\pm0.4 -> 14.0\pm0.6 | Helped |
| Use KNN-SV memory updates rather than reservoir updates (ablation) | ER with random retrieval · Split CIFAR-10, M=0.2k | average accuracy: 26.4\pm1.0 -> 25.9\pm0.6 | Hurt |
| Use ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split Mini-ImageNet, M=1k | average accuracy: 8.7\pm0.4 -> 6.9\pm0.4 | Hurt |
| Use ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split CIFAR-100, M=1k | average accuracy: 11.2\pm0.4 -> 10.1\pm0.3 | Hurt |
| Use ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split CIFAR-10, M=0.2k | average accuracy: 26.4\pm1.0 -> 26.6\pm1.0 | Helped |
| Use mean-ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split Mini-ImageNet, M=1k | average accuracy: 8.7\pm0.4 -> 7.4\pm0.6 | Hurt |
| Use mean-ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split CIFAR-100, M=1k | average accuracy: 11.2\pm0.4 -> 10.8\pm0.3 | Hurt |
| Use mean-ASV retrieval rather than random retrieval, retaining reservoir memory updates (ablation) | ER · Split CIFAR-10, M=0.2k | average accuracy: 26.4\pm1.0 -> 25.8\pm1.0 | Hurt |
Closest papers by meaning
- Non-exemplar Online Class-Incremental Continual Learning via Dual-Prototype Self-Augment and Refinement (AAAI 2024 · similarity 0.88)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.87)
- Autoencoder-Based Hybrid Replay for Class-Incremental Learning (ICML 2025 · similarity 0.86)
- RanDumb: Random Representations Outperform Online Continually Learned Representations (NeurIPS 2024 · similarity 0.86)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · similarity 0.86)
- Summarizing Stream Data for Memory-Constrained Online Continual Learning (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.