AI Descartes
Repository:
Description
Open source package for accelerated symbolic discovery of fundamental laws, the system integrates data-driven symbolic regression with knowledge-based automated reasoning machinary. The repo include 3 main components:
- symbolic-regression - symbolic regression module that offers various methods for formation of symbolic hypotheses based on given data, super-set of permissible operators, and other desired prefrences
- symbolic-discovery-reasoning - differential dynamic reasoning module that recieves background theory and set of hypotheses and qualifies as for their degree of derivability
- experimental-design - given a set of cadidate hypotheses, of various funcitonal forms and / or parameterization, proposes experiments to establish which hypothesis is more likely
For more details please refer to the following mansucirpts:
- https://arxiv.org/abs/2109.01634 AI Descartes: Combining Data and Theory for Derivable Scientific Discovery, C Cornelio, S Dash, V Austel, T Josephson, J Goncalves, K Clarkson, N Megiddo, B El Khadir, L Horesh
- https://arxiv.org/abs/2006.06813 Symbolic Regression using Mixed-Integer Nonlinear Optimization, V Austel, C Cornelio, S Dash, J Goncalves, L Horesh, T Josephson, N Megiddo
Main Contributors
Sanjeeb Dash, Joao Goncalves, Cristina Cornelio, Ken Clarkson, Lior Horesh