Negatively Correlated Search

TL;DR

Negatively Correlated Search (NCS) is a novel evolutionary algorithm that promotes diversity by modeling and encouraging negatively correlated search behaviors, leading to improved performance on complex non-convex optimization problems.

Contribution

This paper introduces NCS, a new EA that explicitly models and promotes negatively correlated search behaviors to enhance diversity and optimization effectiveness.

Findings

NCS achieved the best overall performance on 20 multimodal problems.

NCS outperformed state-of-the-art methods in a linear antenna array synthesis case study.

NCS demonstrates competitive results in non-convex optimization tasks.

Abstract

Evolutionary Algorithms (EAs) have been shown to be powerful tools for complex optimization problems, which are ubiquitous in both communication and big data analytics. This paper presents a new EA, namely Negatively Correlated Search (NCS), which maintains multiple individual search processes in parallel and models the search behaviors of individual search processes as probability distributions. NCS explicitly promotes negatively correlated search behaviors by encouraging differences among the probability distributions (search behaviors). By this means, individual search processes share information and cooperate with each other to search diverse regions of a search space, which makes NCS a promising method for non-convex optimization. The cooperation scheme of NCS could also be regarded as a novel diversity preservation scheme that, different from other existing schemes, directly promotes diversity at the level of search behaviors rather than merely trying to maintain diversity among candidate solutions. Empirical studies showed that NCS is competitive to well-established search methods in the sense that NCS achieved the best overall performance on 20 multimodal (non-convex) continuous optimization problems. The advantages of NCS over state-of-the-art approaches are also demonstrated with a case study on the synthesis of unequally spaced linear antenna arrays.