/eeg_rsa_opt

Primary LanguagePython

EEG-CNN Comparison

  1. Compute EEG RDMs using perform_all_analyses_new.py with the arguments set accordingly. See run_scripts/compute_rdms.sbatch.
  2. Optimize linear combinations of VGG19 weights so that the RDM obtained from each model layer maximally correlates with the EEG RDM. To do this, run compute_optimal_correlation_flex.py with the arguments set accordingly. See run_scripts/optimize_linear_combo.sh.

Note that VGG19 features for pool1, pool2, pool3, pool4, pool5, fc2 model layers must be pre-computed (with the 72 stimuli) and saved in layer_features/.