Neural Networks and Collaborative Filtering

This paper discusses several methods and experiments in Collaborative filtering. Every set of methods has its own advantages and disadvantages and provides better results under different circumstances, which is why these methods are being combined, in the recent days. Each of these methods has been analyzed to see if they are appropriate for a particular phase of the customer decision process. We have discussed the conventional Collaborative filtering (CF) based methods that use factor models, matrix factorization as well as the newly proposed methods like content aware CF, Collaborative deep learning (CDL), recurrent Neural Networks (RNNs), session-based recommendation, neural CF (using a multi-layer perceptron) and Collaborative topic regression (CTR). We have also discussed how the well-known cold-start problem is dealt with in the recent methods.