Large-scale semi-supervised learning with online spectral graph sparsification

2026-04-29

AI Summary

This work addresses the high computational and memory costs inherent in large-scale semi-supervised learning by proposing Sparse-HFS, an algorithm that introduces online spectral graph sparsification to this domain for the first time. By integrating techniques from graph signal processing, Sparse-HFS achieves significant reductions in both time and space complexity while preserving model accuracy. Specifically, it operates with $O(n \,\text{polylog}(n))$ space and $O(m \,\text{polylog}(n))$ time, yielding near-linear time and sublinear space complexity. This efficiency enables effective semi-supervised learning on extremely large graphs that were previously intractable due to resource constraints.

Abstract

We introduce Sparse-HFS, a scalable algorithm that can compute solutions to SSL problems using only O(n polylog(n)) space and O(m polylog(n)) time.

Problem

• semi-supervised learning
• large-scale
• spectral graph sparsification
• scalability
• space efficiency

Innovation

• semi-supervised learning
• spectral graph sparsification
• scalable algorithm
• online learning
• graph-based SSL