Research map / Label-efficient and robust learning / Recognition and long-tail
KAKURENBO: Adaptively Hiding Samples in Deep Neural Network Training
Thao Nguyen Truong, Balazs Gerofi, Edgar Josafat Martinez-Noriega, François Trahay, Mohamed Wahib
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 |
|---|---|---|---|
| increase max hidden fraction from 0.2 to 0.4 (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (B) | top-1 accuracy: 76.11 -> 75.62 | Hurt |
| increase max hidden fraction from 0.2 to 0.4 (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (B) | training time (sec): 61723 -> 57582 | Improved |
| remove move-back (v1110 -> v1010) (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (A) | top-1 accuracy: 72.81 -> 72.96 | Improved |
| add reducing fraction (RF) (v1100 -> v1110) (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (A) | top-1 accuracy: +0.59% | Improved |
| add learning rate adjustment (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (A) | top-1 accuracy: 0.46% to 0.83% | Improved |
| use all strategies vs hiding examples only (ablation) | KAKURENBO · ImageNet-1K ResNet-50 (A) | top-1 accuracy: 73.6 vs baseline 73.68 | Improved |
| increase prediction confidence threshold tau (ablation) | KAKURENBO · ImageNet-1K | accuracy / speedup | Mixed |
| hide samples randomly instead of by loss (ablation) | Random hiding · CIFAR-100 WRN-28-10 | top-1 accuracy: 77.20 -> 76.82 | Hurt |
| increase global batch size from 1024 to 8192 (ablation) | KAKURENBO-0.4 · ImageNet-1K ResNet-50 (A) | top-1 accuracy: 73.60 -> 72.84 | Hurt |
| use KAKURENBO instead of Selective Backprop | KAKURENBO · ImageNet-1K | top-1 accuracy: 3.5% degradation for SB vs +0.2 for KAKURENBO | Improved |
| use ISWR instead of baseline | ISWR · ImageNet-1K | training time: 41% overhead | Hurt |
| use FORGET on WideResNet | FORGET · CIFAR-100 | training time: 46.1% increase | Hurt |
Closest papers by meaning
- Not All Samples Are Created Equal: Deep Learning with Importance Sampling (ICML 2018 · similarity 0.86)
- Lightweight Dataset Pruning without Full Training via Example Difficulty and Prediction Uncertainty (ICML 2025 · similarity 0.84)
- Improving the Training of Data-Efficient GANs via Quality Aware Dynamic Discriminator Rejection Sampling (CVPR 2025 · similarity 0.84)
- Hiding Images in Diffusion Models by Editing Learned Score Functions (CVPR 2025 · similarity 0.83)
- Swift Sampler: Efficient Learning of Sampler by 10 Parameters (NeurIPS 2024 · similarity 0.82)
- Real-Fake: Effective Training Data Synthesis Through Distribution Matching (ICLR 2024 · similarity 0.82)
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.
