Evolving Continuous Optimisers from Scratch

Abstract

This work uses genetic programming to explore the space of continuous optimisers, with the goal of discovering novel ways of doing optimisation. In order to keep the search space broad, the optimisers are evolved from scratch using Push, a Turing-complete, general-purpose, language. The resulting optimisers are found to be diverse, and explore their optimisation landscapes using a variety of interesting, and sometimes unusual, strategies. Significantly, when applied to problems that were not seen during training, many of the evolved optimisers generalise well, and often outperform existing optimisers. This supports the idea that novel and effective forms of optimisation can be discovered in an automated manner. This paper also shows that pools of evolved optimisers can be hybridised to further increase their generality, leading to optimisers that perform robustly over a broad variety of problem types and sizes.

Neuroevolution for Parameter Adaptation in Differential Evolution

“…As opposed to the classical approach, whereby new algorithms are developed by hand and based on the expert knowledge of researchers, several studies have made attempts to automate the process of searching for optimization algorithms and techniques. In particular, some studies addressed the problem of designing new algorithms directly [11,12] with advanced genetic programming techniques, such as PushGP. Other studies, for example, Refs.…”