Research map / Continual learning / Online and replay-based
Lifelong Compression Mixture Model via Knowledge Relationship Graph
Fei Ye, Adrian G. Bors
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 |
|---|---|---|---|
| Use diversity-aware sample selection (DivSS) instead of sliding-window selection (ablation) | LGMM · Split MNIST | classification accuracy: 96.81 ± 0.12 -> 96.95 ± 0.13 | Improved |
| Use diversity-aware sample selection (DivSS) instead of sliding-window selection (ablation) | LGMM · Split CIFAR10 | classification accuracy: 50.91 ± 0.25 -> 53.71 ± 0.19 | Improved |
| Use diversity-aware sample selection (DivSS) instead of sliding-window selection (ablation) | LGMM · Split CIFAR100 | classification accuracy: 25.65 ± 0.16 -> 26.03 ± 0.16 | Improved |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST | classification accuracy: 96.95 ± 0.13 -> 96.16 ± 0.11 | Hurt |
| Add component discarding to DivSS (ablation) | LGMM · Split CIFAR10 | classification accuracy: 53.71 ± 0.19 -> 50.12 ± 0.23 | Hurt |
| Add component discarding to DivSS (ablation) | LGMM · Split CIFAR100 | classification accuracy: 26.03 ± 0.16 -> 25.24 ± 0.17 | Hurt |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST-SVHN | classification accuracy: 65.29 -> 63.78 | Hurt |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST-CIFAR10 | classification accuracy: 64.36 -> 62.15 | Hurt |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST-SVHN | number of components: 18 -> 13 | Improved |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST-SVHN | number of parameters: 175M -> 126M | Improved |
| Add component discarding to DivSS (ablation) | LGMM · Split MNIST-SVHN | testing time per sample: 0.0064 -> 0.0048 | Improved |
| Use a larger expansion threshold λ (ablation) | LGMM · Split MNIST | classification accuracy | Hurt |
Closest papers by meaning
- Online Task-Free Continual Learning via Dynamic Expansionable Memory Distribution (CVPR 2025 · similarity 0.87)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.87)
- Model Sensitivity Aware Continual Learning (NeurIPS 2024 · similarity 0.86)
- Dynamic Expansion Diffusion Learning for Lifelong Generative Modelling (AAAI 2025 · similarity 0.86)
- CMT: A Memory Compression Method for Continual Knowledge Learning of Large Language Models (AAAI 2025 · similarity 0.86)
- Self-Expansion of Pre-trained Models with Mixture of Adapters for Continual Learning (CVPR 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.
