This is a 2-page condensed abstract of the paper published at [1]. Bayesian Matrix Factorization Matrix factorization is a common machine learning technique for recommender systems, like books for Amazon or movies for Net ix [2]. The Bayesian Matrix Factorization (BMF) variant is especially powerful because it produces good results and is relatively robust against over tting. As sketched in Figure 1, the idea of these methods is to approximate the usermovie rating matrix R as a product of two low-rank matrices U and V such that R U V . In this way U and V are constructed from the known ratings in R, which is usually very sparsely lled. The recommendations can be made from the approximation U V which is dense. High-Performance Matrix Factorization with SMURFF While BMF is powerful, it is computationally very intensive and thus more challenging to implement for large datasets. In this work we present SMURFF a high-performance featurerich framework to compose and construct di erent Bayesian matrix-factorization methods, namely:
N
M R =
K U x
M
V ≪ ≪ R is very sparse
K Copyright c 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Authors Suppressed Due to Excessive Length
{ BPMF, the basic version [
Applications of SMURFF The framework has been used very successfully in
drug discovery. Here a key problem is the identi cation of candidate molecules
that a ect proteins associated with diseases [
Getting SMURFF SMURFF is available as open-source [