Research map / Continual learning / Class-incremental
Loss Decoupling for Task-Agnostic Continual Learning
Yan-Shuo Liang, Wu-Jun Li
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 |
|---|---|---|---|
| add LODE to DER++ | DER++ · Seq-CIFAR10 (buffer 500) | accuracy: 73.29 -> 75.45 | Improved |
| add LODE to DER++ | DER++ · Seq-CIFAR100 (buffer 500) | accuracy: 42.08 -> 46.31 | Improved |
| add LODE to DER++ | DER++ · Seq-TinyImageNet (buffer 500) | accuracy: 19.28 -> 21.15 | Improved |
| add LODE to ESMER | ESMER · Seq-CIFAR100 (buffer 500) | accuracy: 52.37 -> 55.06 | Improved |
| add LODE to ER | ER · Seq-CIFAR100 (buffer 500) | accuracy: 27.69 -> 41.52 | Improved |
| add LODE to ER-ACE-like ER, buffer 5120 | ER · Seq-CIFAR10 (buffer 5120) | accuracy: 83.64 -> 83.73 | No clear effect |
| set beta1=beta2=C (remove decoupling) (ablation) | LODE (DER++) · Seq-CIFAR100 | accuracy: 46.31 -> 42.08 | Hurt |
| set beta1=beta2=rho|Cn|/|Co| (remove decoupling) (ablation) | LODE (DER++) · Seq-CIFAR100 | accuracy: 46.31 -> 37.49 | Hurt |
| exchange beta1 and beta2 (ablation) | LODE (DER++) · Seq-CIFAR100 | accuracy: 46.31 -> 40.79 | Hurt |
| set beta1=beta2=C (remove decoupling) (ablation) | LODE (ESMER) · Seq-CIFAR100 | accuracy: 55.06 -> 52.37 | Hurt |
| set beta1=beta2=rho|Cn|/|Co| (ablation) | LODE (ESMER) · Seq-CIFAR100 | accuracy: 55.06 -> 45.64 | Hurt |
| set beta1=beta2=C on Seq-CIFAR10 (ablation) | LODE (ESMER) · Seq-CIFAR10 | accuracy: 74.53 -> 73.08 | Hurt |
Closest papers by meaning
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.89)
- Pareto Continual Learning: Preference-Conditioned Learning and Adaption for Dynamic Stability-Plasticity Trade-off (AAAI 2025 · similarity 0.88)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025 · similarity 0.88)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.88)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · similarity 0.88)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 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.
