Parallel and Distributed Approaches for Graph Based Semi-supervised Learning

TL;DR

This paper introduces two parallel and distributed methods for graph-based semi-supervised learning: an affine map iteration and an MCMC sampling approach, both effective in classifying nodes with low error and adaptable to new data.

Contribution

It presents novel parallel and distributed algorithms for semi-supervised learning on graphs, including an affine map iteration and an MCMC sampling method, with theoretical and practical validation.

Findings

Nodes classified with very small error

Sampling algorithm effectively tracks new nodes

Distributed approach easily implemented on multiple processors

Abstract

Two approaches for graph based semi-supervised learning are proposed. The firstapproach is based on iteration of an affine map. A key element of the affine map iteration is sparsematrix-vector multiplication, which has several very efficient parallel implementations. The secondapproach belongs to the class of Markov Chain Monte Carlo (MCMC) algorithms. It is based onsampling of nodes by performing a random walk on the graph. The latter approach is distributedby its nature and can be easily implemented on several processors or over the network. Boththeoretical and practical evaluations are provided. It is found that the nodes are classified intotheir class with very small error. The sampling algorithm's ability to track new incoming nodesand to classify them is also demonstrated.