[Site Map] You are here --- > Home > Research

aEUS - A Parameter-free Method for High-dimensional Continuous Optimization


Optimization methods that are designed to solve large-scale, high-dimensional problems are generally difficult to implement, require parameter tuning and are computationally expensive. In this paper, a new adaptive variant of the Enhanced Unidimensional Search (EUS) method is proposed to solve this type of problems. The proposed method, called aEUS, is easy to understand and code and requires no parameter tuning. It is tested on 11 scalable benchmark functions using 50, 100, 200, 500 and 1000-dimension versions of each function. Our method is compared with 17 state-of-the-art methods. The results show that our simple approach performs very well compared to the other generally more sophisticated methods.

Source code

aEUS source code is available for download:


aEUS is also downloadable as part of the HeurisTest Java application. This application can be run independently on the ISDA'09 Benchmark. Several other methods are also implemented: EUS, PSO, Tabu Search, Nelder Mead, etc.

Supplementary material