ReLU neural networks of polynomial size for exact maximum flow computation

Hertrich, Christoph ORCID: 0000-0001-5646-8567 and Sering, Leon (2023) ReLU neural networks of polynomial size for exact maximum flow computation. In: Del Pia, Alberto and Kaibel, Volker, (eds.) Integer Programming and Combinatorial Optimization - 24th International Conference, IPCO 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Springer Science and Business Media Deutschland GmbH, pp. 187-202. ISBN 9783031327254

Text (ReLU Neural Networks of Polynomial Size for Exact Maximum Flow Computation) - Accepted Version Download (686kB)

• Scopus publication
• Author

Abstract

This paper studies the expressive power of artificial neural networks with rectified linear units. In order to study them as a model of real-valued computation, we introduce the concept of Max-Affine Arithmetic Programs and show equivalence between them and neural networks concerning natural complexity measures. We then use this result to show that two fundamental combinatorial optimization problems can be solved with polynomial-size neural networks. First, we show that for any undirected graph with n nodes, there is a neural network (with fixed weights and biases) of size O(n3) that takes the edge weights as input and computes the value of a minimum spanning tree of the graph. Second, we show that for any directed graph with n nodes and m arcs, there is a neural network of size O(m2n2) that takes the arc capacities as input and computes a maximum flow. Our results imply that these two problems can be solved with strongly polynomial time algorithms that solely uses affine transformations and maxima computations, but no comparison-based branchings.

Item Type:	Book Section
Additional Information:	© 2023, The Author(s) under exclusive license to Springer Nature Switzerland AG
Divisions:	Mathematics
Subjects:	Q Science > QA Mathematics Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Date Deposited:	28 Jul 2023 09:57
Last Modified:	02 Dec 2024 22:39
URI:	http://eprints.lse.ac.uk/id/eprint/119860

Actions (login required)

View Item