KADAL | Kriging代理模型Python工具箱推荐

简介

用于分析、设计优化和探索的克里金法 (Kriging for Analysis, Design optimization, And expLoration, KADAL) 是由万隆理工学院 (Institut Teknologi Bandung, ITB) 的流动诊断研究小组（Flow Diagnostics Research Group）[1]开发的 Python 程序，其中包含贝叶斯优化工具集合，包括各种代理模型方法、采样技术和优化方法。目前，该程序正在开发中，尚未在 Pypi 上提供。该程序的覆盖范围仍然仅限于：

• 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)
• Reliability Analysis
  AK-MCS
• Sensitivity Analysis
  Sobol indices

Required packages

需要以下第三方库：

• numpy
• scipy
• matplotlib
• sobolsampling
• scikit-learn
• cma

KADAL 已在 Python 3.6.1 上测试

Quick Examples

演示代码位于主文件夹中：

• KrigDemo.py is a demo script for creating surrogate model.
• 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

原始程序由 Pramudita Satria Palar、Kemas Zakaria 和 Ghifari Adam Faza 编写，并由空气动力学研究组 ITB 维护。

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

特别感谢：

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

参考资料

[1] Their Lab’s Homepage: https://flowdiagnostics.ftmd.itb.ac.id