Machine Learning Tools for Mathematical Discovery

Explore machine learning applications in mathematical research through this seminar lecture from Harvard CMSA's New Technologies in Mathematics series. Discover various ML tools currently available for finding interesting constructions to mathematical problems, including simple reinforcement learning setups and PatternBoost methodology. Focus primarily on LLM-based tools such as FunSearch and AlphaEvolve, examining their capabilities for mathematical discovery. Analyze the advantages and disadvantages of different ML approaches to determine optimal methods for specific mathematical problems. Learn from collaborative research involving experts from Google DeepMind and leading academic institutions, gaining insights into the intersection of artificial intelligence and mathematical research methodology.