A new paper in Neuroimage applies deep neural nets to the classification of schizophrenia vs controls from resting state fMRI data. It outperforms other approaches and also extracts some interesting hierarchical features. The paper discusses pre-training and sparsity control in some detail.

Deep neural network with weight sparsity control and pre-training extracts hierarchical features and enhances classification performance patterns of schizophrenia

Deep nets for schizophrenia rs-fMRI classification