edna_glm

preprocessing

• edit experiment paths in main() in features.py and song_utils.py
• run in terminal with

  conda activate env
  python features.py
  python song_utils.py
  

• the outputs of these files are hdf5 files with kinematic behavior features or song features

GLM

• glm.py
  • decide stimulus history window in frames
  • change experiment paths and savepaths in main()
  • run in terminal with

    conda activate env
    python glm.py
    

• the output is png files of the filters and a results.csv containing:
  • pCor = percent correct
  • logloss
  • filter_norms
  • f1_score
  • deviance (calculated by 2*sklearn.metrics.log_loss(y, model.predict_proba(x), normalize=False))