Research map / Continual learning / Online and replay-based
Natural continual learning: success is a journey, not (just) a destination
Ta-Chu Kao, Kristopher T Jensen, Gido Martijn van de Ven, Alberto Bernacchia, Guillaume Hennequin
Read the paperCheck 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 |
|---|---|---|---|
| NCL vs other methods in task-incremental setting | NCL · split MNIST and split CIFAR-100 | average test accuracy | Improved |
| NCL vs other methods in domain-incremental setting | NCL · split MNIST and split CIFAR-100 | average test accuracy | Improved |
| NCL vs OWM in class-incremental setting | NCL · split MNIST and split CIFAR-100 | average test accuracy | Hurt |
| NCL vs SI, online EWC and KFAC in class-incremental setting | NCL · split MNIST and split CIFAR-100 | average test accuracy | Improved |
| unit Gaussian prior without hyperparameter optimization (NCL no-prior) (ablation) | NCL · split MNIST and split CIFAR-100 | average test accuracy | No clear effect |
| NCL vs OWM and DOWM | NCL · stimulus-response RNN tasks | mean loss | Improved |
| NCL, OWM and DOWM vs KFAC with lambda=1 | NCL · stimulus-response RNN tasks | mean loss | Improved |
| increase KFAC lambda to between 100 and 1000 (ablation) | KFAC · stimulus-response RNN tasks | mean loss | Improved |
| NCL vs DOWM and KFAC lambda=1 on SMNIST | NCL · SMNIST 15 tasks | mean classification error | Improved |
| optimize lambda for KFAC (ablation) | KFAC · SMNIST 15 tasks | mean classification error | Improved |
| Laplace-DOWM projection approximation in place of NCL Fisher approximation (ablation) | NCL · SMNIST, first task | classification error | Improved |
| Laplace-DOWM projection approximation in place of NCL Fisher approximation (ablation) | NCL · SMNIST, later tasks | classification error | Hurt |
Closest papers by meaning
- On the Theory of Continual Learning with Gradient Descent for Neural Networks (ICML 2026 · similarity 0.91)
- Continual Learning through Control Minimization (ICML 2026 · similarity 0.90)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · similarity 0.89)
- Learning Continually by Spectral Regularization (ICLR 2025 · similarity 0.89)
- A Statistical Theory of Regularization-Based Continual Learning (ICML 2024 · similarity 0.88)
- Temporal-Difference Variational Continual Learning (NeurIPS 2025 · 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.
