Scalable tensor methods for nonuniform hypergraphs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Aksoy, Sinan G., Amburg, Ilya, Young, Stephen J.
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909160151449600
author Aksoy, Sinan G.
Amburg, Ilya
Young, Stephen J.
author_facet Aksoy, Sinan G.
Amburg, Ilya
Young, Stephen J.
contents While multilinear algebra appears natural for studying the multiway interactions modeled by hypergraphs, tensor methods for general hypergraphs have been stymied by theoretical and practical barriers. A recently proposed adjacency tensor is applicable to nonuniform hypergraphs, but is prohibitively costly to form and analyze in practice. We develop tensor times same vector (TTSV) algorithms for this tensor which improve complexity from $O(n^r)$ to a low-degree polynomial in $r$, where $n$ is the number of vertices and $r$ is the maximum hyperedge size. Our algorithms are implicit, avoiding formation of the order $r$ adjacency tensor. We demonstrate the flexibility and utility of our approach in practice by developing tensor-based hypergraph centrality and clustering algorithms. We also show these tensor measures offer complementary information to analogous graph-reduction approaches on data, and are also able to detect higher-order structure that many existing matrix-based approaches provably cannot.
format Preprint
id arxiv_https___arxiv_org_abs_2306_17825
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Scalable tensor methods for nonuniform hypergraphs
Aksoy, Sinan G.
Amburg, Ilya
Young, Stephen J.
Numerical Analysis
Machine Learning
Social and Information Networks
Combinatorics
Physics and Society
05C65, 15A69, 05C50, 05C85
While multilinear algebra appears natural for studying the multiway interactions modeled by hypergraphs, tensor methods for general hypergraphs have been stymied by theoretical and practical barriers. A recently proposed adjacency tensor is applicable to nonuniform hypergraphs, but is prohibitively costly to form and analyze in practice. We develop tensor times same vector (TTSV) algorithms for this tensor which improve complexity from $O(n^r)$ to a low-degree polynomial in $r$, where $n$ is the number of vertices and $r$ is the maximum hyperedge size. Our algorithms are implicit, avoiding formation of the order $r$ adjacency tensor. We demonstrate the flexibility and utility of our approach in practice by developing tensor-based hypergraph centrality and clustering algorithms. We also show these tensor measures offer complementary information to analogous graph-reduction approaches on data, and are also able to detect higher-order structure that many existing matrix-based approaches provably cannot.
title Scalable tensor methods for nonuniform hypergraphs
topic Numerical Analysis
Machine Learning
Social and Information Networks
Combinatorics
Physics and Society
05C65, 15A69, 05C50, 05C85
url https://arxiv.org/abs/2306.17825