New solver accelerates graph $p$-Laplacian semi-supervised learning

Researchers have developed a novel solver for graph $p$-Laplacian semi-supervised learning that achieves near-linear time complexity. This new method addresses limitations of existing solvers, particularly at higher values of $p$ where the system can become ill-conditioned. By employing a continuation approach in $p$ and a damped chord-Newton method, the solver maintains well-conditioned systems, leading to significant speedups and reduced memory usage compared to direct factorization methods. The approach demonstrates improved performance on large-scale graph families and benchmark datasets like MNIST, outperforming standard quadratic ($p=2$) methods. AI

IMPACT This research offers a more efficient method for graph-based semi-supervised learning, potentially enabling larger and more complex datasets to be processed.

RANK_REASON The cluster contains a research paper detailing a new algorithmic approach for a specific machine learning task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Chord-Newton
• Dirichlet
• Ferrovie Calabro Lucane
• Graph $p$-Laplacian
• Laplace operator
• MNIST database
• Nadler-Srebro-Zhou
• semi-supervised learning
• SuiteSparse

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →