Learning advanced mathematical computations from examples
This addresses the challenge of automating complex mathematical analysis for researchers and engineers, though it appears incremental as it applies existing transformer methods to a new domain.
The paper tackled the problem of training neural networks to learn advanced mathematical computations, such as differential system properties, from examples without built-in knowledge, achieving near-perfect prediction of qualitative characteristics and good approximations of numerical features.
Using transformers over large generated datasets, we train models to learn mathematical properties of differential systems, such as local stability, behavior at infinity and controllability. We achieve near perfect prediction of qualitative characteristics, and good approximations of numerical features of the system. This demonstrates that neural networks can learn to perform complex computations, grounded in advanced theory, from examples, without built-in mathematical knowledge.