Edited code

code/challenge_baseline_model/02_encoding_model_training/train_encoding_utils.py

@@ -190,7 +190,13 @@ def load_stimulus_features(args, movie_split_names, movie_split_samples):
 				# using the corresponding stimulus feature sample minus the
 				# hrf_delay
 				elif mod == 'language':
-					idx = s + args.excluded_samples_start - args.hrf_delay
+					# In case there are no language features for the fMRI
+					# sample of interest minus the hrf_delay, model the fMRI
+					# sample using the first language feature sample
+					if s < args.hrf_delay:
+						idx = args.excluded_samples_start
+					else:
+						idx = s + args.excluded_samples_start - args.hrf_delay
 					# In case there are fewer language feature samples than
 					# fMRI samples minus the hrf_delay, use the last language
 					# feature sample available

code/challenge_baseline_model/03_encoding_model_testing/test_encoding_utils.py

@@ -157,7 +157,13 @@ def load_stimulus_features(args, subject, fmri_samples):
 				# spanning several samples, only model each fMRI sample using
 				# the corresponding stimulus feature sample minus the hrf_delay
 				elif mod == 'language':
-					idx = s - args.hrf_delay
+					# In case there are no language features for the fMRI
+					# sample of interest minus the hrf_delay, model the fMRI
+					# sample using the first language feature sample
+					if s < args.hrf_delay:
+						idx = 0
+					else:
+						idx = s - args.hrf_delay
 					# In case there are fewer language feature samples than
 					# fMRI samples minus the hrf_delay, use the last language
 					# feature sample available