Transformer---EEGxAI

Transformer-based prediction and explanation pipeline for EEG-derived features

EEG Transformer + SHAP

This project builds a machine learning pipeline for classifying states of consciousness from EEG-derived features using a transformer-based model. Predictions are explained using SHAP for interpretability in clinical settings.

Project Structure

• data/: raw or synthetic EEG-derived features
• notebooks/: development notebooks
• src/: source code for training and explaining models
• plots/: figures generated for SHAP and performance analysis

Goals

• Train a transformer or strong ML model on EEG graph features
• Use SHAP to explain predictions at global and local levels
• Provide interpretable results for clinical insight

Status

🔧 In progress — dataset generation and pipeline implementation underway.