What is it about?
This paper describes recent advances in the machine learning method SISSO, as implemented in the new code SISSO++. It describes various updates to how SISSO handles the input features for a given model and illustrates their impact with toy examples. Finally it discusses how these tools can be used for the physical sciences.
Featured Image
Photo by Chris Ried on Unsplash
Why is it important?
This work is important as it begins to introduce "grammar rules" into symbolic regression. Symbolic regression tries to find a mathematical equation, e.g. P = (x + z)^2, to approximate a target property, P. This work creates new ways to limit the possible expressions to ones that are physically meaningful.
Perspectives
This work represents a culmination of my time implementing and updating they SISSO methodology. Fundamentally, the work tries to make symbolic regression more physically meaningful by focusing on expressions that are possible. This work not only speeds up the application of the code, but also improves its accuracy and usability.
Thomas Purcell
Fritz Haber Institute of the Max Planck Society
Read the Original
This page is a summary of: Recent advances in the SISSO method and their implementation in the SISSO++ code, The Journal of Chemical Physics, September 2023, American Institute of Physics,
DOI: 10.1063/5.0156620.
You can read the full text:
Resources
Contributors
The following have contributed to this page
