Neural Discovery in Mathematics: Do Machines Dream of Colored Planes?

Konrad Mundinger, Max Zimmer, Aldo Kiem, Christoph Spiegel, Sebastian Pokutta

International Conference on Machine Learning 2025 · Oral

This talk, presented by researchers from the IOL Research Lab in Berlin, delves into the fascinating intersection of machine learning and pure mathematics, presenting a compelling case study of AI-driven scientific discovery. The central theme revolves around leveraging advanced ML tools, particularly implicit representations and gradient-based optimization, to guide mathematical intuition and uncover novel structures for long-standing open problems. The researchers frame their work as a method to assist in the discovery of "extremal structures," a common challenge in mathematics where the goal is to find an optimal configuration or arrangement under specific constraints.

AI review

An honest and technically competent application of implicit neural representations (SIREN-style networks) to the Hadwiger-Nelson coloring problem, producing one verifiable improvement — sub-4% plane removal for 5-colorings — and new off-diagonal constructions. The methodology is clean, the experimental validation is credible, and the framing is appropriately modest about what ML is actually doing here (hypothesis generation, not proving). The work sits at an interesting intersection of combinatorial geometry and continuous optimization, but the theoretical depth is limited: the neural…