Research map / Label-efficient and robust learning / Semi-supervised learning
Deep Growing Learning
Guangcong Wang, Xiaohua Xie, Jianhuang Lai, Jiaxuan Zhuo
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 the selection sub-network and growing sub-network (full DGL) (ablation) | fixed four-block supervised baseline · CIFAR-10 (4000 labels) | test error (%): 35.38 -> 17.56 | Improved |
| add the selection sub-network and growing sub-network (full DGL) (ablation) | fixed four-block supervised baseline · CIFAR-10 (10000 labels) | test error (%): 24.05 -> 12.77 | Improved |
| add the selection sub-network and growing sub-network (full DGL) (ablation) | fixed four-block supervised baseline · CIFAR-10 (20000 labels) | test error (%): 16.70 -> 10.41 | Improved |
| add the selection sub-network and growing sub-network (full DGL) (ablation) | fixed four-block supervised baseline · CIFAR-10 (40000 labels) | test error (%): 12.39 -> 8.07 | Improved |
| add network growing instead of a fixed one-block network (ablation) | DGL with self-training · CIFAR-10 (4000 labels) | test error (%): 30.27 -> 17.56 | Improved |
| add network growing instead of a fixed two-block network (ablation) | DGL with self-training · CIFAR-10 (4000 labels) | test error (%): 26.98 -> 17.56 | Improved |
| add network growing instead of a fixed three-block network (ablation) | DGL with self-training · CIFAR-10 (4000 labels) | test error (%): 30.16 -> 17.56 | Improved |
| add network growing instead of a fixed four-block network (ablation) | DGL with self-training · CIFAR-10 (4000 labels) | test error (%): 33.43 -> 17.56 | Improved |
| add pseudo-label selection (self-training) to growing network (ablation) | DGL · CIFAR-10 (4000 labels) | test error (%): performance drops by 8.96% | Improved |
| raise labeled set from 500 to 4,000 examples | DGL · CIFAR-10 | test error (%): performs significantly better from 500 to 4,000, slight increase from 4,000 to 40,000 | Improved |
| vary confidence threshold alpha in range 86-99% (ablation) | DGL · CIFAR-10 (4000 labels) | test error (%): less than 18% test error for alpha in 86-99% | No clear effect |
Closest papers by meaning
- Realistic Evaluation of Deep Semi-Supervised Learning Algorithms (NeurIPS 2018 · similarity 0.88)
- Towards Realistic Semi-Supervised Learning (ECCV 2022 · similarity 0.86)
- DP-SSL: Towards Robust Semi-supervised Learning with A Few Labeled Samples (NeurIPS 2021 · similarity 0.86)
- Revisiting Semi-Supervised Learning in the Era of Foundation Models (NeurIPS 2025 · similarity 0.85)
- Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection (NeurIPS 2024 · similarity 0.85)
- Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning (ICML 2024 · similarity 0.84)
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.
