[SYSTEMDS-1437] Implement and scale Factorization Machines using SystemML

• Type: Task
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: SystemML 1.2
• Component/s: Algorithms
• Labels: factorization_machines scalability
• Assignee: Janardhan Pulivarthi
• Reporter: Imran Younus
• Created: 25/Mar/17 00:19
• Updated: 18/May/18 03:49
• Resolved: 18/May/18 03:49

Description

Factorization Machines have gained popularity in recent years due to their effectiveness in recommendation systems. FMs are general predictors which allow to capture interactions between all features in a features matrix. The feature matrices pertinent to the recommendation systems are highly sparse. SystemML's highly efficient distributed sparse matrix operations can be leveraged to implement FMs in a scalable fashion. Given the closed model equation of FMs, the model parameters can be learned using gradient descent methods.

Implementation of factorization machines, as described in the paper, as a core fm.dml module to support

• Regression
• Binary classification
• Ranking

We'll showcase the scalability of SystemML, with an end-to-end recommender system. Possibly, we could integrate some other algorithms to build a state-of-the-art recommender system.

paper: http://www.algo.uni-konstanz.de/members/rendle/pdf/Rendle2010FM.pdf

Mentors: iyounus, nakul02, dusenberrymw

Issue Links

• relates to: SYSTEMDS-2102 Vectorize gradients for Factorization Machines function
• relates to: SYSTEMDS-2103 Add unit tests for the Factorization Machines core layer module