Department of Computer Engineering
CS 590 SEMINAR
Matrix Factorization with Stochastic Gradient Descent for Recommender Systems
Ömer Faruk Aktulum
Computer Engineering Department
Matrix factorization is an efficient technique used for disclosing latent features of real world data. It finds its application in areas such as text mining, image analysis, social network and more recently and popularly in recommendation systems. Alternating least squares, stochastic gradient descent (SGD) and coordinate descent are among the methods used for factorization. SGD-based factorization has proven to be the most successful one among them after Netflix and KDDCup competitions where the winners' algorithms relied on methods based on SGD. Parallelization of SGD then became a hot topic and studied extensively in the literature. Distributed memory parallelizations include works such as DSGD, ASGD and NOMAD and shared memory parallelizations include works such as FPSGD, HOGWILD and cuMF. These approaches mostly focus on load balancing issues without paying much attention to communication overheads. In this study, we investigate models and methods which address not only load balance but also communication overheads.
DATE: 24 April, 2017, Monday @ 16:30