2019/05/20 by Olivier Gouvert, Gouvert, Olivier, Thomas Oberlin +3
Computer Science · #Advanced Graph Neural Networks #Advanced Image and Video Retrieval Techniques #Data Management and Algorithms #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Recommender Systems and Techniques
paper · pdf · doi:10.48550/arxiv.1905.13128
openalex publication_date 2019/05/20 · openalex created_date 2022/09/05 · openalex updated_date 2026/07/28
Count data are often used in recommender systems: they are widespread (song\nplay counts, product purchases, clicks on web pages) and can reveal user\npreference without any explicit rating from the user. Such data are known to be\nsparse, over-dispersed and bursty, which makes their direct use in recommender\nsystems challenging, often leading to pre-processing steps such as\nbinarization. The aim of this paper is to build recommender systems from these\nraw data, by means of the recently proposed compound Poisson Factorization\n(cPF). The paper contributions are three-fold: we present a unified framework\nfor discrete data (dcPF), leading to an adaptive and scalable algorithm; we\nshow that our framework achieves a trade-off between Poisson Factorization (PF)\napplied to raw and binarized data; we study four specific instances that are\nrelevant to recommendation and exhibit new links with combinatorics.\nExperiments with three different datasets show that dcPF is able to effectively\nadjust to over-dispersion, leading to better recommendation scores when\ncompared with PF on either raw or binarized data.\n