smt-sgp-paper

Repo related to the paper 'A Python Toolbox for Data-Driven Aerodynamic Modeling using Sparse Gaussian Processes'

Citation

Valayer, H.; Bartoli, N.; Castaño-Aguirre, M.; Lafage, R.; Lefebvre, T.; López-Lopera, A.F.; Mouton, S. A Python Toolbox for Data-Driven Aerodynamic Modeling Using Sparse Gaussian Processes. Aerospace 2024, 11, 260. https://doi.org/10.3390/aerospace11040260

SMT version

SMT 2.3.0 was used to get the results.

pip install smt==2.3.0 

Repository Content

• Notebooks

  • Analytic test case : sparse_gp_analytic.ipynb
  • Wind Tunnel test case: sparse_gp_wtdata.ipynb (Warning: wind tunnel data are not publicly available)

• FIG_ANALYTIC directory contains figures generated from analytic test case notebook

• FIG_WT directory contains figures generated from wind tunnel test case notebook

• wtdata_results directory contains csv results generated by following scripts from wind tunnel data

• sparse_gp_wtdata.py generates sgp_wtdata_results_M\<value\>.csv and deals with all output variable, all sparse methods, all inducing methods for a given M nb of inducing points

• sparse_gp_wtdata_czc_nkmeans.py generates sgp_wtdata_results_czc_nkmeans.csv and deals with the CZC output variable, normalized kmeans and various M values

• sparse_gp_wtdata_noseed_10times.py generates sgp_wtdata_results_noseed_M50_10times.csv and deals with all output variable, all sparse methods, all inducing methods by repeating ten times with no random seed

SMT Sparse GP tutorial

See SGP analytic notebook