Reviews: Geometric Matrix Completion with Recurrent Multi-Graph Neural Networks

This paper generalizes a recent NIPS 2016 paper [10] by allowing convolutional neural network (CNN) models to work on multiple graphs. It extracts local stationary patterns from signals defined on the graphs simultaneously. In particular, it is applied to recommender systems by considering the graphs defined between users and between items. The multi-graph CNN model is followed by a recurrent neural network (RNN) with long short-term memory (LSTM) cells to complete the score matrix. Strengths of the paper: * The proposed deep learning architecture is novel for solving the matrix completion problem in recommender systems with the relationships between users and the relationships between items represented as two graphs.