Research map / Continual learning / Class-incremental
Incremental Learning in Online Scenario
Jiangpeng He, Runyu Mao, Zeman Shao, Fengqing Zhu
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Combine the baseline classifier with NCM representation learning and switch from NCM after the baseline performs better for five consecutive blocks | Online learning from scratch on Food-101 · Food-101, 20/30/40/50 classes | online accuracy: 62.81% -> 70.90%; 56.53% -> 64.32%; 54.35% -> 62.31%; 51.39% -> 57.83% | Improved |
| Use modified cross-distillation with accommodation ratio and two-step learning instead of original cross-distillation | Online incremental learning from a trained model · Food-101, new classes | online accuracy: 54.35% -> 64.78%; 52.62% -> 62.25%; 46.30% -> 61.53% | Improved |
| Use modified cross-distillation with accommodation ratio and two-step learning instead of original cross-distillation | Online incremental learning from a trained model · Food-101, old classes | online accuracy: 22.83% -> 61.01%; 22.41% -> 60.00%; 20.53% -> 53.43% | Improved |
| Use modified cross-distillation with accommodation ratio and two-step learning instead of original cross-distillation | Online incremental learning from a trained model · Food-101, new classes | test accuracy: 70.97% -> 64.00%; 71.56% -> 61.87%; 66.62% -> 56.31% | Hurt |
| Use modified cross-distillation with accommodation ratio and two-step learning instead of original cross-distillation | Online incremental learning from a trained model · Food-101, old classes | test accuracy: 41.77% -> 70.32%; 42.25% -> 69.90%; 40.82% -> 65.65% | Improved |
| Use modified cross-distillation instead of cross-entropy or original cross-distillation (ablation) | Online incremental learning from a trained model · Food-101 | online accuracy | Improved |
| Update the exemplar set using features of new observations of old classes (ablation) | Online incremental learning from a trained model · Food-101 | overall online accuracy | Improved |
| Update the exemplar set using features of new observations of old classes (ablation) | Online incremental learning from a trained model · Food-101, old classes | online accuracy | Improved |
Closest papers by meaning
- Class-Incremental Learning: Survey and Performance Evaluation on Image Classification (TPAMI 2022 · similarity 0.90)
- End-to-End Incremental Learning (ECCV 2018 · similarity 0.90)
- Always Be Dreaming: A New Approach for Data-Free Class-Incremental Learning (ICCV 2021 · similarity 0.90)
- Incremental Learning of Object Detectors Without Catastrophic Forgetting (ICCV 2017 · similarity 0.89)
- Class-Incremental Learning: A Survey (TPAMI 2024 · similarity 0.88)
- Advancing Cross-domain Discriminability in Continual Learning of Vision-Language Models (NeurIPS 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.
