Note

This repo is no longer active and just for experiments, newer version is available on https://github.com/flowdiagnosticsitb/KADAL

KADAL

Kriging for Analysis, Design optimization, And expLoration (KADAL) is a Python program developed by Flow Diagnostics Research Group from Institut Teknologi Bandung (ITB) that contains collections of Bayesian Optimization tools including various surrogate modeling methods, sampling techniques, and optimization methods. Currently, the program is under development and not implemented yet as a module for Python 3. Also, the coverage of the program are still limited to:

• Kriging
  • Ordinary Kriging
  • Regression Kriging
  • Polynomial Kriging
  • Composite Kernel Kriging
  • Kriging with Partial Least Square
• Bayesian Optimization
  • Unconstrained Single-Objective Bayesian Optimization (Expected Improvement)
  • Unconstrained Multi-Objective Bayesian Optimization (ParEGO, EHVI)
• Test Cases
  • Branin (Single-Objective)
  • Sasena (Single-Objective)
  • Styblinski-Tang (Single-Objective)
  • Schaffer (Multi-Objective)

Required packages

KADAL depends on these modules: numpy, scipy, sk-learn, and pycma.

• pycma: https://github.com/CMA-ES/pycma

KADAL has been tested on Python 3.6.1

Quick Examples

The demo codes are available in the main folder.

• KrigDemo.py is a demo script for creating surrogate model for Branin test function.
• MOBOdemo.py is a demo script for performing unconstrained multi-objective optimization for Schaffer test function.
• SOBOdemo.py is a demo script for performing unconstrained single objective optimization for Branin test function.

Contact

The original program was written by Pramudita Satria Palar, Kemas Zakaria, Ghifari Adam Faza, and maintained by Aerodynamics Research Group ITB.

e-mail: pramsp@ftmd.itb.ac.id

Special thanks to:

• Timothy Jim (Tohoku University)
• Potsawat Boonjaipetch (Tohoku University)