Clinamen 0.2.0a3

Purpose

Clinamen: Covariance-matrix-adaptation evolution strategy for Localized Impurities with unsupervised Machine learning Exploration of Non-elementary structures.

Clinamen is a software package developed in order to assist in the discovery of low-energy defective configurations in first-principles atomistic calculations.

The approach is based on a modified version of the covariance-matrix-adaptation evolution-strategy (CMA-ES), a state-of-the-art evolutionary algorithm for continuous-parameter multimodal optimization. In addition, the approach employs an unsupervised machine learning model for identifying additional local minima explored during the run of the evolutionary algorithm. A supervised Gaussian-process regressor trained on-the-fly during the evolutionary process can also be used in order to reduce the number of first-principles calculations.

Clinamen is written mostly in Python and employs Cython and C extension modules for an increased computational speed.

Installation

The code is available on the PyPi index. The easiest way to install Clinamen is through pip:

pip install clinamen

We suggest to create a virtual environment in which the Clinamen code can be installed.

The requirements for a successful installation are:

• Python version 3.8 or newer;

• NumPy version 1.18 or newer;

• SciPy version 1.5 or newer;

• Pandas version 1.0 or newer;

• scikit-learn version 0.23 or newer;

• Cython version 0.29 or newer;

• mpi4py version 3.0 or newer;

• h5py version 2.10 or newer;

• Atomic Simulation Environment version 3.19 or newer;

• PyTorch version 1.7 or newer;

• GPyTorch version 1.3 or newer;

• DScribe version 0.4 or newer.

It might be that some packages cannot be installed directly with the Clinamen package using pip (for example PyTorch installation could fail in this way). For this reason these packages have to be installed manually before installing Clinamen with pip.