Research map / Label-efficient and robust learning / Semi-supervised learning
SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification
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 |
|---|---|---|---|
| Add Pair Loss to SimPLE (ablation) | SimPLE · CIFAR-100, 10000 labels, K=2 | Top-1 test accuracy: 69.07% -> 70.82% | Helped |
| Add Pair Loss to SimPLE with K=7 (ablation) | SimPLE · CIFAR-100, 10000 labels, K=7 | Top-1 test accuracy: 69.94% -> 73.04% | Helped |
| Increase weak augmentations K from 2 to 7 (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 70.82% -> 73.04% | Helped |
| Use RandAugment instead of fixed augmentations (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 67.91% -> 70.82% | Helped |
| Lower confidence threshold to 0.75 (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 70.82% -> 71.96% | No clear effect |
| Lower similarity threshold to 0.7 (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 70.82% -> 70.85% | No clear effect |
| Use Pair Loss weight 50 instead of 150 (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 70.82% -> 71.34% | No clear effect |
| Use Pair Loss weight 250 instead of 150 (ablation) | SimPLE · CIFAR-100, 10000 labels | Top-1 test accuracy: 70.82% -> 71.42% | No clear effect |
| Use L2 distance for Pair Loss instead of the Bhattacharyya-based distance (ablation) | SimPLE · Experiments reported in text | Test accuracy: 0.53% | Hurt |
| Add Pair Loss | SimPLE · CIFAR-10 and SVHN | Top-1 test accuracy | No clear effect |
| Use SimPLE instead of the strongest reported baseline | Semi-supervised classification · CIFAR-100, 10000 labels, WRN 28-8 | Top-1 test accuracy: 77.40% -> 78.11% | Helped |
| Use SimPLE instead of the strongest reported baseline | Semi-supervised classification · Mini-ImageNet, 4000 labels, WRN 28-2 | Top-1 test accuracy: 60.50% -> 66.55% | Helped |
Closest papers by meaning
- All Labels Are Not Created Equal: Enhancing Semi-Supervision via Label Grouping and Co-Training (CVPR 2021 · similarity 0.92)
- Barely-Supervised Learning: Semi-supervised Learning with Very Few Labeled Images (AAAI 2022 · similarity 0.92)
- FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020 · similarity 0.91)
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · similarity 0.90)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.89)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · 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.