Model-centric computation abstractions in machine learning applications

Abstract

We categorize parallel machine learning applications into four types of computation models and propose a new set of model-centric computation abstractions. This work sets up parallel machine learning as a combination of training data-centric and model parameter-centric processing. The analysis uses Latent Dirichlet Allocation (LDA) as an example, and experimental results show that an efficient parallel model update pipeline can achieve similar or higher model convergence speed compared with other work.