The document describes a method for collaborative subspace clustering using a deep neural network. The network contains an encoder, a self-expressive layer to learn the affinity matrix C, and a decoder. The network is trained end-to-end by minimizing a loss function containing terms for subspace clustering and collaborative learning between the affinity matrix C and a classifier's output affinity matrix. The loss encourages C to be more confident in identifying points from the same class compared to the classifier.