“AI for Mathematics” (AI4Math) in the general sense refers to using AI, machine learning, and neural networks to solve problems in pure and applied mathematics, including AI4PDE which finds approximate solutions to partial differential equations.

“AI for Mathematics” (AI4Math) in the narrow sense refers to the neuro-symbolic approach to generate candidates for mathematical proofs, until being accepted by formal verification systems like Lean, Coq, and Isabelle, using Large Language Models (LLMs), and (possibly “human-in-the-loop”) Reinforcement Learning (RL).

Optimization: Stochastic Gradient Descent Backpropagation Neuro-symbolic AI4Math