Neural Collaborative Filtering vs. Matrix Factorization Revisited

Abstract

Embedding based models have been the state of the art in collaborative filtering for over a decade. Traditionally, the dot product or higher order equivalents have been used to combine two or more embeddings, e.g., most notably in matrix factorization. In recent years, it was suggested to replace the dot product with a learned similarity e.g. using a multilayer perceptron (MLP). This approach is often referred to as neural collaborative filtering (NCF). In this work, we revisit the experiments of the NCF paper that popularized learned similarities using MLPs. First, we show that with a proper hyperparameter selection, a simple dot product substantially outperforms the proposed learned similarities. Second, while a MLP can in theory approximate any function, we show that it is non-trivial to learn a dot product with an MLP. Finally, we discuss practical issues that arise when applying MLP based similarities and show that MLPs are too costly to use for item recommendation in production environments while dot products allow to apply very efficient retrieval algorithms. We conclude that MLPs should be used with care as embedding combiner and that dot products might be a better default choice. CCS CONCEPTS • Information systems → Recommender systems; Collaborative filtering; • Computing methodologies → Learning from implicit feedback; Neural networks; Factorization methods.

Neural Explicit Factor Model Based on Item Features for Recommendation Systems

“…The dot product is widely used to calculate similarity. Unless the dataset is large or the embedding dimension is very small, the inner product is a suitable choice as a measure for embeddings [4].…”

“…Recommender systems play a pivotal role in alleviating information overload [1] and are widely used in applications such as Web search, e-commerce, and entertainment. Current recommendation systems mainly rely on collaborative filtering (CF) [2]- [4] to provide useful results. CF is a technique based on a user's historical data (e.g., browsing history, comments) and has attracted much attention for recommendation systems due to its benefits.…”

Movie Recommendation through Multiple Bias Analysis

“…R( f ) is the set of rated scores belonging to feature f , r(v, j) is the rated score for item j of user v, ω(u, f ) is the weight of feature f of user u, and ω(i, f ) is the weight of feature f of item i. The range of the values of the feature weight ω is −1 to 1, and the value that is computed through Equation 26is used in Equations (28) and (29). In GF, ω is optimized to compute how close the feature f was to the popular bias.…”

“…The λ used in Equations (26) and (27) is the learning rate, and |GF(s, u, i)| and |PF(s, u, i)| are the absolute values of score. The value computed with Equation 26learns about the ω used in GF through Equations (28) and (29), and the value computed with Equation (27) learns about the ω in PF through Equation (29). Because PF is computed using only IF(i), only Equation (29) is used to learn the ω.…”

“…Because PF is computed using only IF(i), only Equation (29) is used to learn the ω. The TF( f ) that is used in Equations (28) and (29) means the output of the upper sub-feature to which feature f belongs. If feature f is an action, the output of TF( f ) will become Genre.…”