Finding Increasingly Large Extremal Graphs with AlphaZero and Tabu Search

Finding Increasingly Large Extremal Graphs with AlphaZero and Tabu Search

Abbas Mehrabian, Ankit Anand, Hyunjik Kim, Nicolas Sonnerat, Matej Balog, Gheorghe Comanici, Tudor Berariu, Andrew Lee, Anian Ruoss, Anna Bulanova, Daniel Toyama, Sam Blackwell, Bernardino Romera Paredes, Petar Veličković, Laurent Orseau, Joonkyung Lee, Anurag Murty Naredla, Doina Precup, Adam Zsolt Wagner

Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence
Main Track. Pages 6985-6993. https://doi.org/10.24963/ijcai.2024/772

This work proposes a new learning-to-search benchmark and uses AI to discover new mathematical knowledge related to an open conjecture of Erdos (1975) in extremal graph theory. The problem is to find graphs with a given size (number of nodes) that maximize the number of edges without having 3- or 4-cycles. We formulate this as a sequential decision-making problem and compare AlphaZero, a neural network-guided tree search, with tabu search, a heuristic local search method. Using either method, by introducing a curriculum---jump-starting the search for larger graphs using good graphs found at smaller sizes---we improve the state-of-the-art lower bounds for several sizes. We also propose a flexible graph-generation environment and a permutation-invariant network architecture for learning to search in the space of graphs.
Keywords:
Search: S: Search and machine learning
Multidisciplinary Topics and Applications: MTA: Other
Search: S: Local search
Search: S: Combinatorial search and optimisation