Collaborative Memory Network for Recommendation Systems

Abstract

Recommendation systems play a vital role to keep users engaged with personalized content in modern online platforms. Deep learning has revolutionized many research fields and there is a recent surge of interest in applying it to collaborative filtering (CF). However, existing methods compose deep learning architectures with the latent factor model ignoring a major class of CF models, neighborhood or memory-based approaches. We propose Collaborative Memory Networks (CMN), a deep architecture to unify the two classes of CF models capitalizing on the strengths of the global structure of latent factor model and local neighborhood-based structure in a nonlinear fashion. Motivated by the success of Memory Networks, we fuse a memory component and neural attention mechanism as the neighborhood component. The associative addressing scheme with the user and item memories in the memory module encodes complex user-item relations coupled with the neural attention mechanism to learn a user-item specific neighborhood. Finally, the output module jointly exploits the neighborhood with the user and item memories to produce the ranking score. Stacking multiple memory modules together yield deeper architectures capturing increasingly complex user-item relations. Furthermore, we show strong connections between CMN components, memory networks and the three classes of CF models. Comprehensive experimental results demonstrate the effectiveness of CMN on three public datasets outperforming competitive baselines. Qualitative visualization of the attention weights provide insight into the model's recommendation process and suggest the presence of higher order interactions.

BHIN2vec

“…Using the node pairs and edge, we create the edge embedding vectors by applying Hadamard function to two node vectors in the training set [6]. We train a logistic regression classifier using the edge embedding vectors and evaluate the hit rate at 10 for the recommendation task, which is widely used in the recommendation field [5,9,22]. Given an edge (v i , v j ) in test set, we sample 99 nodes in which the sampled node type is same as the type of v j .…”

DBRec

“…historical items rated by users) to compensate the interaction function. Ebesu et al [7] propose an attention to aggregate neighboring users, and jointly exploit the neighborhoods with user-item interactions to derive the recommendation. However, one major shortcoming of those models is that, they are mainly based on the interaction data, and suffer from data sparseness.…”

“…In [7], the authors combine latent factor model and neighborhoodbased structure, and propose an attention mechanism to find similar users based on the specific user and item. However, they simply aggregate similar users for estimating the rating scores, therefore the neighboring users are not sufficiently leveraged for bridging unobserved user-item pairs.…”

“…The similarities between z u i and all the group representations are measured with the inner product operation, and the resulted logits are transformed into a posterior probability vector with a softmax function, as shown in Eq. (7).…”

“…It uses outer product to transform latent vectors of a user-item pair into two-dimensional map, and applies convolutional and pooling layers to model deep user-item interactions. • CMN [7]. It identify similar neighboring users with an attention mechanism based on the specific user-item pair, and jointly exploits the neighborhood state and user-item interactions to derive recommendation.…”