- Two main folders, one for each dataset
IMAGES_BLOCK_MASK
: Block-design experiment, with voxels filtering according to probability of belonging to gray matter. This is the dataset used in the paper.IMAGES_EVENT_MASK
: Event-design experiment, with voxels filtering according to probability of belonging to gray matter.SOLVERS
: contains the optimization solvers.UTILITIES
: containes auxiliary functions, such as for thresholding solutions and theNIFTI
toolbox for Matlab.
- Each dataset folder contains subfolders for data and for each of the models, e.g.,
LASSO
,ENET
,SLAP
. - Each of the models' folders contain the Matlab functions to run the numerical experiment:
MODEL_val.m
runs the folds of the internal LOSO loop.MODEL_train.m
runs the folds of the external LOSO loop, using the results obtained viaMODEL_val.m
.- Both functions require as input argument the number of the fold to execute, as a string, to allow naïve parallel processing.
compute_results_MODEL.m
computes all the performance metrics and saves them intoRESULTS_TRAIN\MODEL_esults.mat
.
wangdeqing / neurosparse Goto Github PK
View Code? Open in Web Editor NEWThis project forked from lucabaldassarre/neurosparse