Research map / Label-efficient and robust learning / Semi-supervised learning
Mixture of Expert/Imitator Networks: Scalable Semi-Supervised Learning Framework
Read the paperCheck 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 trained IMNs | LSTM · Elec, IMDB, Rotten, RCV1 | error rate (%): 10.09 -> 8.83; 10.98 -> 10.04; 26.47 -> 24.93; 14.14 -> 12.31 | Helped |
| Add trained IMNs | LM-LSTM · Elec, IMDB, Rotten, RCV1 | error rate (%): 5.72 -> 5.48; 7.25 -> 6.51; 16.80 -> 15.91; 8.37 -> 7.53 | Helped |
| Add trained IMNs | ADV-LM-LSTM · Elec, IMDB, Rotten, RCV1 | error rate (%): 5.38 -> 5.14; 6.58 -> 6.07; 15.73 -> 13.98; 7.89 -> 7.51 | Helped |
| Replace trained IMNs with random vectors (ablation) | LSTM, LM-LSTM, and ADV-LM-LSTM · Elec, IMDB, Rotten, RCV1 | error rate (%): LSTM: 10.09 -> 9.87, 10.98 -> 10.75, 26.47 -> 27.27, 14.14 -> 14.04 | Mixed |
| Use IMN window sizes 1,2,3,4 rather than only window size 4 (ablation) | ADV-LM-LSTM · Elec | error rate (%): 5.23 -> 5.14 | Helped |
| Remove the window-size-1 IMN (ablation) | ADV-LM-LSTM + IMN (c_i=1,2,3,4) · Elec | error rate (%): 5.14 -> 5.18 | Hurt |
| Remove the window-size-1 and window-size-2 IMNs (ablation) | ADV-LM-LSTM + IMN (c_i=1,2,3,4) · Elec | error rate (%): 5.14 -> 5.26 | Hurt |
| Add an IMN with window size 5 (ablation) | ADV-LM-LSTM + IMN · Elec, IMDB | error rate (%): 5.14 -> 5.12; 6.07 -> 6.00 | Helped |
| Increase the IMN window size (ablation) | ADV-LM-LSTM + IMN · Elec, IMDB, Rotten, RCV1 | error rate (%) | Helped |
| Double the word embedding dimension (ablation) | ADV-LM-LSTM · IMDB | error rate (%): degraded by 0.08 points | Hurt |
| Double the word embedding dimension (ablation) | ADV-LM-LSTM · Elec | error rate (%): unchanged | No clear effect |
| Use ELMo embeddings instead of the RNN language-model initialization | LSTM · IMDB | error rate (%): 8.67% versus 7.25 | Hurt |
Closest papers by meaning
- Improved Graph Contrastive Learning for Short Text Classification (AAAI 2024 · similarity 0.86)
- A Square Peg in a Square Hole: Meta-Expert for Long-Tailed Semi-Supervised Learning (ICML 2025 · similarity 0.85)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.85)
- Toward a Unified Statistical Theory of Unsupervised Pretraining and Supervised Neural Knowledge Graph Learning (NeurIPS 2026 · similarity 0.85)
- Three Heads Are Better than One: Complementary Experts for Long-Tailed Semi-supervised Learning (AAAI 2024 · similarity 0.85)
- MoDE: CLIP Data Experts via Clustering (CVPR 2024 · similarity 0.85)
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.