Fig. 7From: PASNet: pathway-associated sparse deep neural network for prognosis prediction from high-throughput dataTraining of PASNet. (a) Weights and biases are randomly initialized. Connections between the gene layer and the pathway layer are determined by biological pathway databases, and the remaining layers are considered as fully-connected in this step. (b) A sub-network is randomly selected using a dropout technique and trained. (c) Sparse coding optimizes the sparsity of connections in the sub-networkBack to article page