Research map / Continual learning / Online and replay-based
Theory on Forgetting and Generalization of 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 |
|---|---|---|---|
| Give more weight to fresher correlated old tasks | TRGP · PMNIST | ACC (%): 96.34 -> 96.75 | Helped |
| Give more weight to fresher correlated old tasks | TRGP · PMNIST | BWT (%): -0.8 -> -0.46 | Helped |
| Give more weight to fresher correlated old tasks | TRGP · Split CIFAR-100 | ACC (%): 74.46 -> 75.31 | Helped |
| Give more weight to fresher correlated old tasks | TRGP · Split CIFAR-100 | BWT (%): -0.9 -> 0.13 | Helped |
| Increase model parameter count in the overparameterized regime | Continual learning in overparameterized linear models · Continual learning tasks | forgetting | Helped |
| Increase model parameter count in the overparameterized regime | Continual learning in overparameterized linear models · Continual learning tasks | generalization error | Helped |
| Make tasks more similar | Continual learning in overparameterized linear models · Continual learning tasks | generalization error | Helped |
| Make tasks less similar in the four-task scenario | Continual learning in overparameterized linear models · Four-task setup | forgetting | Helped |
| Diversify tasks early and place dissimilar tasks adjacently | Continual learning · Task ordering | forgetting | Helped |
Closest papers by meaning
- The Joint Effect of Task Similarity and Overparameterization on Catastrophic Forgetting — An Analytical Model (ICLR 2024 · similarity 0.93)
- Unlocking the Power of Rehearsal in Continual Learning: A Theoretical Perspective (ICML 2025 · similarity 0.91)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.91)
- Towards Understanding Catastrophic Forgetting in Two-layer Convolutional Neural Networks (ICML 2025 · similarity 0.91)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.91)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.91)
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.