How krxgu/eeg_research-work works
Main Components
• Data Sources: Include components such as the BCI Competition IV dataset and possibly custom EEG datasets. Label these as “EEG Data/BCI Competition Dataset” or similar.
• Preprocessing/Exploration: Although not explicitly broken out in separate directories, note that each notebook probably contains its own data preprocessing steps. Indicate that raw EEG data is fed into each experiment module.
• Experiment Modules/Models:
– BCI_Competition_IV: A module that directly uses the BCI Competition IV data.
– ChronoNet: Split this into two sub-components:
▪ ChronoNet Implementation - Keras
▪ ChronoNet Implementation - PyTorch
As two technical implementations of the same model, highlighting the dual framework approach.
– DL_Classification: A deep learning based approach dedicated to EEG classification.
– Epilepsy_Classification: Specialized module focused on classifying epilepsy from EEG signals.
– ML_classification: Where classical machine learning techniques are applied to EEG data.
– ResNet-Classification: An experimental module using (likely) a variant of the ResNet architecture applied to EEG.
Relationships and Data Flow
• If there are any common pre-/post-processing elements (e.g., data normalization or feature extraction steps mentioned within the notebooks), denote those as shared components or as part of the flow before the data splits into the experimental pipelines.
• Finally, indicate that the outputs (model predictions, performance metrics, visualizations) from each module converge into an “Evaluation/Results” component.
Architectural Patterns and Design Principles
• Modular structure: each experiment is decoupled and can be developed or run independently.
• That there is a separation by framework – for instance, ChronoNet has clear splits between Keras and PyTorch implementations. Indicate these branches distinctly.