A semi-supervised classification RBM with an improved fMRI representation algorithm

•An unsupervised feature learning algorithm named HRBM is used for RBM to make the fMRI feature representation learned sparse.•A semi-supervised classification RBM for fMRI with a joint tuning algorithm based on the improved HRBM, namely semi-HRBM is proposed.•Compared with the supervised models, the performance of Semi-HRBM was significantly improved.•Our HRBM has satisfactory feature representation capabilities and better performance for multiple classification tasks.•Our Semi-HRBM classification model improves the average accuracy of the four-classification task by 7.68%, and improves the average F1 score of each visual stimulus task by 8.90%.