From ZuCo 1.0 and 2.0: NR and TSR tasks
On UZH server:
ZuCo 1.0: methlab/NLP/Ce-ETH/FirstLevel_concat_unfold_correctlyMergedSacc_
ZuCo 2.0: methlab/NLP/Ce_ETH/2019/FirstLevelV2_concat_unfold_correctlyMergedSacc
On spaceML:
noraho@spaceml3:/mnt/ds3lab-scratch/noraho/datasets/zuco/zuco1_preprocessed_sep2020
noraho@spaceml3:/mnt/ds3lab-scratch/noraho/datasets/zuco/zuco2_preprocessed_sep2020
Set parameters in config.py
Feature extraction:
extract_features.py
data_loading_helpers.py
Main script:
classify_nr_trs.py
Train & test classifier for each subject individually
classify_nr_tsr_cross.py
Leave-one-out cross-subject models: train on all-1 subjects, test on left out subject
classify_sessions.py
Classify recording sessions (this uses SR data from ZuCo 1)
classify_blocks.py
Classify recording blocks from ZuCO 2
classify_subects.py
Subject classification
classify_nr_trs_WordFixOnly.py
Features include only data during fixation.
Description: Section 6.3 Fixation ablation - Figure 20
Set parameters in config.py
Feature extraction:
data_loading_helpers.py
eeg_extractor.py
gaze_extractor.py
Best to do feature extraction once separately and save features for faster processing
tune_eeg_model_single.py
tune_gaze_model_single.py
Train & test classifier for each subject individually
tune_gaze_model_cross.py
tune_eeg_model_cross.py
Leave-one-out cross-subject models: train on all-1 subjects, test on left out subject