Research map / Continual learning / Online and replay-based
A Neural Dirichlet Process Mixture Model for Task-Free Continual Learning
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 |
|---|---|---|---|
| Use CN-DPM instead of Reservoir | Task-free continual learning · Split-MNIST | accuracy (%): 85.69 -> 93.23 | Helped |
| Use CN-DPM instead of iid-online | Task-free continual learning · MNIST-SVHN | accuracy (%): 95.24 -> 94.46 | Hurt |
| Use CN-DPM instead of iid-online | Task-free continual learning · Split-CIFAR100 | accuracy (%): 20.46 -> 20.10 | Hurt |
| Use CN-DPM instead of Reservoir | Task-free continual learning · Split-CIFAR10 (1 Epoch) | accuracy (%): 44.00 -> 45.21 | Helped |
| Use CN-DPM instead of iid-online | Task-free continual learning · Split-MNIST (Gen.) | bits/dim: 0.2156 -> 0.2110 | Helped |
| Extend task presentation from 1 Epoch to 10 Epochs | Reservoir · Split-MNIST | accuracy (%): 85.69 ± 0.48 -> 78.82 ± 0.71 | Hurt |
| Extend task presentation from 1 Epoch to 10 Epochs | CN-DPM · Split-MNIST | accuracy (%): 93.23 ± 0.09 -> 94.39 ± 0.04 | Helped |
| Revisit tasks across 10 scenario passes instead of presenting each task for 10 consecutive epochs | Reservoir · Split-MNIST | accuracy (%): 78.82 ± 0.71 -> 92.06 ± 0.11 | Helped |
| Use task labels to select experts during training | CN-DPM · Split-MNIST classification | accuracy (%): 93.70 ± 0.07 -> 93.81 ± 0.07 | Helped |
| Use CN-DPM instead of Reservoir | Task-free continual learning without clear task boundaries · Fuzzy Split-MNIST | accuracy (%): 88.64 ± 0.48 -> 93.22 ± 0.07 | Helped |
| Prune redundant experts 3 and 6 (ablation) | CN-DPM · Split-MNIST | accuracy (%): 87.07% -> 86.01% | Hurt |
Closest papers by meaning
- Task-Free Continual Learning via Online Discrepancy Distance Learning (NeurIPS 2022 · similarity 0.89)
- Probing Representation Forgetting in Supervised and Unsupervised Continual Learning (CVPR 2022 · similarity 0.89)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.88)
- Optimal Task Order for Continual Learning of Multiple Tasks (ICML 2025 · similarity 0.88)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.88)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.88)
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.