Sampling for Approximate Bipartite Network Projection

TL;DR

This paper introduces a novel sampling algorithm for efficiently estimating the similarity matrix in bipartite network projections, enabling scalable analysis of large streaming data with high accuracy.

Contribution

The paper proposes a fixed-size, unbiased sampling method that preferentially samples high-similarity node pairs in bipartite streams, improving estimation efficiency.

Findings

Achieves about 1% weighted relative error with only 10% sampling.

Effectively estimates high-similarity edges in real-world bipartite graphs.

Demonstrates scalability and accuracy in streaming bipartite network analysis.

Abstract

Bipartite networks manifest as a stream of edges that represent transactions, e.g., purchases by retail customers. Many machine learning applications employ neighborhood-based measures to characterize the similarity among the nodes, such as the pairwise number of common neighbors (CN) and related metrics. While the number of node pairs that share neighbors is potentially enormous, only a relatively small proportion of them have many common neighbors. This motivates finding a weighted sampling approach to preferentially sample these node pairs. This paper presents a new sampling algorithm that provides a fixed size unbiased estimate of the similarity matrix resulting from a bipartite graph stream projection. The algorithm has two components. First, it maintains a reservoir of sampled bipartite edges with sampling weights that favor selection of high similarity nodes. Second, arriving edges generate a stream of \textsl{similarity updates} based on their adjacency with the current sample. These updates are aggregated in a second reservoir sample-based stream aggregator to yield the final unbiased estimate. Experiments on real world graphs show that a 10% sample at each stage yields estimates of high similarity edges with weighted relative errors of about 1%.