Generation of Nonlinear Parameter Space

The sure independence screening and sparsifying operator (SISSO) algorithm was published in 2018 by Ouyang et. al. This algorithm inspired this "SISSO-ish" code for application of nonlinear parameters to linear modeling techniques.

What does this code do?

1. Data Cleaning

  1.1. Drop non-numerical columns and any designated to skip

  1.2. Remove colinear parameters from input data set

2. Nonlinear Parameter Generation

  2.1. Perform algebraic transformations on data set

  2.2. Evaluate transformed parameters for p-value improvment from parent parameters

  2.3. Save evaluated parameters and concatenate to original data set.

  2.4. Repeat from 2.1. for iterations (iter=int)

3. Export Transformed and Original Parameters

Example Usage

Publications:

• Original paper: R. Ouyang, S. Curtarolo, E. Ahmetcik, M. Scheffler, and L. M. Ghiringhelli, Phys. Rev. Mater. 2, 083802 (2018).
• Original repository: https://github.com/rouyang2017/SISSO.git

Our Implementation:

• Souza, L. W.§, Miller, B. R. §, Cammarota, R. C., Lo, A., Lopez, I., Shiue, Y. S., Bergstrom, B. S., Fettinger, J., Sigman, M., Shaw, J. T. Deconvoluting Nonlinear Catalyst-Substrate Effects in the Intramolecular Dirhodium-Catalyzed C-H Insertion of Donor/Donor Carbenes Using Data Science Tools. ACS Catalysis, 2024, 14(1), 104–115. https://doi.org/10.1021/acscatal.3c04256