Fig. 1: Autoencoders architecture.
From: Sparsely-connected autoencoder (SCA) for single cell RNAseq data mining
a Sparsely-connected autoencoders (SCA), (b) variational sparsely-connected autoencoders (vSCA), (c) Sparse sparsely-connected autoencoders (SSCA). Gray nodes refer to genes. Gene-level expression profiles for each gene in each cell are used as input and reconstructed as output on the basis of the latent space. The latent space is made of nodes where each node is associated with a transcription factor, a miRNA, a kinase or a functional signature or other biological knowledge. The vertices connecting input/output nodes to latent space are based on experimentally validated biological knowledge.
