Research map / Continual learning / Online and replay-based
Direction Concentration Learning: Enhancing Congruency in Machine Learning
Yan Luo, Yongkang Wong, Mohan Kankanhalli, Qi Zhao
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 DCL constraints | ResNet-50 RMSProp · SALICON 2017 val | NSS: 1.7933 -> 1.8226 | Improved |
| add DCL constraints | ResNet-50 Adam · SALICON 2017 val | NSS: 1.7978 -> 1.8019 | Improved |
| add DCL constraints | DINet Adam · SALICON 2017 val | NSS: 1.8786 -> 1.8857 | Improved |
| add GEM constraints | ResNet-50 Adam · OSIE to MIT1003 | NSS: 2.4064 -> 2.3685 | Hurt |
| add DCL constraints | GEM with MLP · MNIST-R | accuracy: 83.35 -> 84.08 | Improved |
| add DCL constraints | GEM with MLP · MNIST-P | accuracy: 82.44 -> 82.58 | No clear effect |
| add DCL constraints (window 4) | GEM with ResNet · iCIFAR-100 | accuracy: 66.67 -> 67.92 | Improved |
| use larger effective window (4 to 50) (ablation) | DCL MEM with ResNet · iCIFAR-100 | accuracy: 67.92 -> 64.86 | Hurt |
| add DCL | ResNeXt-29 SGD · CIFAR-10 | top-1 error: 3.53 -> 3.33 | Improved |
| add DCL | ResNeXt-29 SGD · CIFAR-100 | top-1 error: 17.30 -> 17.02 | Improved |
| add GEM | ResNeXt-29 SGD · CIFAR-10 | top-1 error: 3.53 -> 7.70 | Hurt |
| add DCL | ResNet-101 SGD · Tiny ImageNet | top-1 error: 17.34 -> 16.89 | Improved |
Closest papers by meaning
- Decoupled Contrastive Learning (ECCV 2022 · similarity 0.85)
- Advancing Cross-domain Discriminability in Continual Learning of Vision-Language Models (NeurIPS 2024 · similarity 0.84)
- Bridging Mini-Batch and Asymptotic Analysis in Contrastive Learning: From InfoNCE to Kernel-Based Losses (ICML 2024 · similarity 0.83)
- DiffusionMTL: Learning Multi-Task Denoising Diffusion Model from Partially Annotated Data (CVPR 2024 · similarity 0.83)
- Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learning (ICML 2025 · similarity 0.83)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.83)
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.
