Across-subject ensemble-learning alleviates the need for large samples for fMRI decoding

Data repository for the paper "Across-subject ensemble-learning alleviates the need for large samples for fMRI decoding" by Himanshu Aggarwal, Liza Al-Shikhley and Bertrand Thirion. A preprint version of this work can be found here: https://hal.science/hal-04636523 This data consists of downsampled 3mm event-wise GLM effect-size maps of fMRI datasets used in the study. Here, N_samples refers to the number of samples per subject, N_subjects to the number of subjects in the cohort, and N_classes to the number of classes in the prediction task. Dataset Task N_samples N_subjects N_classes Stimuli labels Neuromod Visual n-back 50 4 4 images of body/face/place/tools AOMIC Emotion anticipation 61 203 4 negative/neutral emotion images and cue for negative/neutral emotion images Forrest Music genre perception 175 10 5 ambient/country/metal/rocknroll/symphonic music BOLD5000 Image-Net image viewing 332 3 4 images of furniture/vehicle/animal/person RSVP-IBC RSVP languages 360 13 6 type of text: jabberwocky/complex/simple/word list/pseudoword list/consonant strings