Work

Learning advanced mathematical computations from examples

Year: 2020

Type: preprint

Abstract: 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.

Source: arXiv (Cornell University)

DOI: 10.48550/arxiv.2006.06462

Authors: Charton, François, Hayat, Amaury, Lample, Guillaume

Institutions: Meta (United Kingdom), Meta (United States), Rutgers, The State University of New Jersey, École nationale des ponts et chaussées

Language: English

Topic: Neural Networks and Applications

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W3129421149

Cites: 43

Cited by: 10

Related to: 10