Symbolic Mathematics Finally Yields to Neural Networks
Researchers reframed symbolic math as translation and trained neural networks on tree-structured equations. The system solved many integrals and differential equations quickly, though limits remain.