Research map / Label-efficient and robust learning / Self-supervised and contrastive
EEG-DLite: Dataset Distillation for Efficient Large EEG Model Training
Yuting Tang, Weibang Jiang, Shanglin Li, Yong Li, Chenyu Liu, Xinliang Zhou, Yi Ding, Cuntai Guan
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 |
|---|---|---|---|
| distill pre-training data to 5% with EEG-DLite instead of random 5% | LaBraM-base pre-training · SEED-V | accuracy (%): 34.6 -> 38.6 | Improved |
| distill to 1% with EEG-DLite instead of random | LaBraM-base pre-training · SEED-V | accuracy (%): 26.5 -> 29.7 | Improved |
| distill to 5% with EEG-DLite instead of full data | LaBraM-base pre-training · TUAB / TUEV / MoBI | downstream performance: MoBI PCC 0.538 -> 0.550; TUEV F1 83.1 -> 82.2 at 10% | Mixed |
| use synthetic data generation (M3D) instead of random selection | EEG pre-training distillation · SEED-V | accuracy (%): 34.6 -> 26.9 at 5% | Hurt |
| use SSL autoencoder embeddings instead of PCA for diversity sampling (ablation) | EEG-DLite · SEED-V | accuracy (%): 31.0 -> 38.6 at 5% | Improved |
| add outlier removal (tau = 1%) before diversity sampling (ablation) | EEG-DLite with EEGNet · SEED cross-subject supervised (50% data) | accuracy (%): 54.1 -> 56.6 | Improved |
| add outlier removal (tau = 1%) (ablation) | EEG-DLite with EEGNet · SEED cross-subject supervised (25% data) | accuracy (%): 54.6 -> 55.3 | Improved |
| remove top 0.05% OOD samples before selection (ablation) | EEG-DLite pre-training · SEED-V | accuracy | Improved |
| distill to 25% with EEG-DLite instead of full data | EEGNet supervised · SEED cross-subject | accuracy (%): 54.6 -> 55.3 | Improved |
| increase distillation ratio for random selection from 5% to 25% (ablation) | Random baseline · SEED-V | accuracy (%): 34.6 -> 39.1 | Improved |
| increase distillation ratio from 10% to 25% for EEG-DLite (ablation) | EEG-DLite · TUAB | AUROC (%): 90.8 -> 90.3 | No clear effect |
Closest papers by meaning
- Standing on the Shoulders of Giants: Rethinking EEG Foundation Model Pretraining via Multi-Teacher Distillation (NeurIPS 2026 · similarity 0.90)
- NeuroAtlas: Benchmarking Foundation Models for Clinical EEG and Brain-Computer Interfaces (NeurIPS 2026 · similarity 0.87)
- Pre-Training Graph Contrastive Masked Autoencoders are Strong Distillers for EEG (ICML 2025 · similarity 0.87)
- What do EEG Foundation Models Capture from Human Brain Signals? (NeurIPS 2026 · similarity 0.87)
- BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals (NeurIPS 2025 · similarity 0.87)
- CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding (ICLR 2025 · similarity 0.87)
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.
