Providing easy-to-use open source software in the context of (randomized) numerical blackbox optimization is one of the main goals of the RandOpt team. We have been and are currently involved in the following software projects:

The Covariance Matrix Adaptation Evolution Strategy (CMA-ES):

The CMA-ES (Covariance Matrix Adaptation Evolution Strategy) is an evolutionary algorithm for difficult non-linear non-convex black-box optimisation problems in continuous domain (numerical optimization). The CMA-ES is considered as state-of-the-art in evolutionary computation and has been adopted as one of the standard tools for continuous optimisation in many (probably hundreds of) research labs and industrial environments around the world. The algorithm is a research focus of the team. It is implemented in the cma Python package. The multi-objective COMO-CMA-ES is implemented in the comocma Python package.

CMA-ES Source Code Home Page
CMA-ES on GitHub

The Comparing Continuous Optimizers Platform (COCO):

Developed at Inria since 2007, the COmparing Continuous Optimizers platform (COCO) aims at simplifying the numerical benchmarking of optimization algorithms. It allows to run numerical benchmarking experiments almost automatically on various provided test suites (e.g. unconstrained noiseless or noisy single-objective problems, unconstrained noiseless bi-objective problems) in a large variety of programming languages (C/C++, Python, Java, Matlab/Octave) and to investigate algorithm performances with its Python postprocessing tools. For the latter, 300+ algorithm data sets of the most common optimization algorithms (amongst others stochastic algorithms such as CMA-ES variants, trust-region methods such as NEWUOA, pattern search methods, direct search methods, quasi-Newton methods, …)  are available to the optimization community.

Code on GitHub
Test Suites

Comments are closed.