Skip to content

ChiaraBenini/eeg-ea-robustness-study

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

README

What's included

  • notebook.ipynb — all code, cells 1–14
  • config.yaml — all hyperparameters, read at startup

Before you run

Expected total runtime: 5–7 hours.

Two main bottlenecks:

Dataset (~1 hour). The notebook uses Lee2019_ERP via MOABB, which downloads automatically on first run. This is several GB and can take 30–60 minutes depending on your connection. Once downloaded, MOABB caches it locally. The preprocessed epochs are also cached to data/eeg_cache.pkl — if that file exists, the whole loading step is skipped on re-runs.

Evolutionary strategy (~4–6 hours). The ES runs over multiple seeds with up to ea.max_generations generations each. If you just want a quick check, set n_runs: 1 and max_generations: 100 in config.yaml.

Dependencies

numpy, matplotlib, pyyaml, scikit-learn, mne, moabb, pyriemann

Checkpointing

Results are saved to results/stage2/checkpoints/ after each seed completes, so a crash mid-run doesn't lose prior work. Cell 11 has commented-out code to reload from checkpoints without re-running the EA.

Reproducibility

Everything is seeded from config.yaml. Given the same config and a populated cache, results should be fully reproducible.

About

Evolutionary Neural Architecture Search for Spiking Neural Networks with channel attention (SECA).Evolutionary Neural Architecture Search for Spiking Neural Networks with channel attention (SECA). Optimizes attention placement, time steps, membrane decay, and channel width on CIFAR-10/100. Multi-GPU training via Distrib

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors