pyMaxPro_lite

Simple MaxPro space-filling algorithm implementation in python.

The code utilizes MaxPro criterion described in V. Roshan Joseph, Evren Gul, and Shan Ba (2015), but does not contain fancy functionalities that the original authors implemented in R language.

How to install

You may create or use anaconda environment to control over the packages used in your project.

Create environment:

conda create -y --name custom-env-name python=3.8
conda activate custom-env-name

or activate existing environment:

conda activate custom-env-name

Install prerequisite packages.

Install following packages from conda-forge channel:

SciPy
NumPy

conda install -c conda-forge scipy numpy

Clone and install pyMaxPro_lite.

Clone from github to local storage, and checkout branch if necessary:

git clone https://github.com/yonghoonlee/pyMaxPro_lite.git
cd pyMaxPro_lite
git checkout branch-name

Install pyMaxPro_lite:

python setup.py develop

How to use

Run following simple code:

from pymaxpro_lite.maxpro import maxpro_design
d = maxpro_design()
d.n = 8
d.p = 3
x = d.solve()
print(x)

then you will get:

[[9.18972072e-01 7.31505609e-01 5.91135883e-01]
 [3.60570411e-05 4.26330655e-01 7.78425090e-01]
 [7.42082749e-01 5.24189855e-01 4.38276579e-04]
 [2.25781710e-01 9.99759184e-01 2.80936313e-01]
 [9.99995556e-01 3.41777594e-04 9.99445511e-01]
 [9.25104488e-02 4.34303173e-02 8.66445617e-02]
 [5.62380386e-01 2.01882624e-01 4.36223971e-01]
 [4.53881879e-01 8.68993411e-01 9.36525545e-01]]

Reference

Joseph, V.R., Gul, E., Ba, S., (2015) "Maximum projection designs for computer experiments," Biometrika, 102(2):371–380.