https://arxiv.org/pdf/1511.06939.pdf

2.1 SESSION-BASED RECOMMENDATION

Much of the work in the area of recommender systems has focused on models that work when a user identifier is available and a clear user profile can be built. In this setting, matrix factorization methods and neighborhood models have dominated the literature and are also employed on-line. One of the main approaches that is employed in session-based recommendation and a natural solution to the problem of a missing user profile is the item-to-item recommendation approach (Sarwar et al., 2001; Linden et al., 2003) in this setting an item to item similarity matrix is precomputed from the available session data, that is items that are often clicked together in sessions are deemed to be similar. This similarity matrix is then simply used during the session to recommend the most similar items to the one the user has currently clicked. While simple, this method has been proven to be effective and is widely employed. While effective, these methods are only taking into account the last click of the user, in effect ignoring the information of the past clicks. A somewhat different approach to session-based recommendation are Markov Decision Processes (MDPs) (Shani et al., 2002). MDPs are models of sequential stochastic decision problems. An MDP is defined as a four-tuple hS, A, Rwd, tri where S is the set of states, A is a set of actions Rwd is a reward function and tr is the state-transition function. In recommender systems actions can be equated with recommendations and the simplest MPDs are essentially first order Markov chains where the next recommendation can be simply computed on the basis of the transition probability between items. The main issue with applying Markov chains in session-based recommendation is that the state space quickly becomes unmanageable when trying to include all possible sequences of user selections. The extended version of the General Factorization Framework (GFF) (Hidasi & Tikk, 2015) is capable of using session data for recommendations. It models a session by the sum of its events. It uses two kinds of latent representations for items, one represents the item itself, the other is for representing the item as part of a session. The session is then represented as the average of the feature vectors of part-of-a-session item representation. However, this approach does not consider any ordering within the session.

2.2 DEEP LEARNING IN RECOMMENDERS

One of the first related methods in the neural networks literature where the use of Restricted Boltzmann Machines (RBM) for Collaborative Filtering (Salakhutdinov et al., 2007). In this work an RBM is used to model user-item interaction and perform recommendations. This model has been shown to be one of the best performing Collaborative Filtering models. Deep Models have been used to extract features from unstructured content such as music or images that are then used together with more conventional collaborative filtering models. In Van den Oord et al. (2013) a convolutional deep network is used to extract feature from music files that are then used in a factor model. More recently Wang et al. (2015) introduced a more generic approach whereby a deep network is used to extract generic content-features from any types of items, these features are then incorporated in a standard collaborative filtering model to enhance the recommendation performance. This approach seems to be particularly useful in settings where there is not sufficient user-item interaction information.