Research map / Continual learning / Online and replay-based
Conditional Channel Gated Networks for Task-Aware Continual Learning
Check 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 input-dependent, task-specific channel gates with sparsity and freeze relevant filters | Continual learning model · Split SVHN and Split CIFAR-10, task-incremental | average accuracy: 0.962 -> 0.974; 0.963 -> 0.964 | Helped |
| Use input-dependent, task-specific channel gates with sparsity and freeze relevant filters | Continual learning model · Split MNIST, task-incremental | average accuracy: 0.997 -> 0.997 | No clear effect |
| Rehearse the task classifier using episodic examples | Class-incremental model · Split MNIST and Split SVHN | average accuracy | Helped |
| Rehearse only task prediction rather than the class heads (ablation) | Class-incremental model with generative memory · Split SVHN | accuracy: 0.7847 -> 0.8341 | Helped |
| Rehearse only task prediction rather than the class heads (ablation) | Class-incremental model with generative memory · Split CIFAR-10 | accuracy: 0.6384 -> 0.7006 | Helped |
| Replace class-conditional generators and class-level rehearsal with unconditional task generators and task-level rehearsal (ablation) | Generative-memory model · Split SVHN and Split CIFAR-10 | accuracy: 0.7847 -> 0.8341; 0.6384 -> 0.7006 | Helped |
| Use episodic rather than generative memory (ablation) | Continual learning model · Split SVHN and Split CIFAR-10 | accuracy: 0.8600 vs 0.8341; 0.8132 vs 0.7006 | Helped |
| Mix real current-task examples with generated past-task examples during task-classifier rehearsal (ablation) | Generative-memory task classifier · Class-incremental setting | task prediction performance | Hurt |
| Use generated examples for the current task too, instead of mixing real and fake rehearsal examples (ablation) | Generative-memory task classifier · Class-incremental setting | task prediction performance | Helped |
Closest papers by meaning
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.88)
- Mixture of Experts Meets Prompt-Based Continual Learning (NeurIPS 2024 · similarity 0.87)
- Optimal Task Order for Continual Learning of Multiple Tasks (ICML 2025 · similarity 0.87)
- Towards Understanding Catastrophic Forgetting in Two-layer Convolutional Neural Networks (ICML 2025 · similarity 0.87)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.86)
- Exploiting Task Relationships in Continual Learning via Transferability-Aware Task Embeddings (NeurIPS 2025 · 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.