Non-Gaussian Random Generators in Bacteria Foraging Algorithm for Multiobjective Optimization

TL;DR

This paper explores how non-Gaussian stochastic engines, specifically Weibull and Gamma distributions, influence the performance of the bacteria foraging algorithm in solving a real-world multi-objective optimization problem, providing new insights into stochastic effects.

Contribution

It introduces the use of non-Gaussian random generators in bacteria foraging algorithm and evaluates their impact on multi-objective optimization performance.

Findings

Non-Gaussian generators affect Pareto front quality.

Weibull and Gamma distributions improve exploration.

New AER metric offers better benchmarking.

Abstract

Random generators or stochastic engines are a key component in the structure of metaheuristic algorithms. This work investigates the effects of non-Gaussian stochastic engines on the performance of metaheuristics when solving a real-world optimization problem. In this work, the bacteria foraging algorithm (BFA) was employed in tandem with four random generators (stochastic engines). The stochastic engines operate using the Weibull distribution, Gamma distribution, Gaussian distribution and a chaotic mechanism. The two non-Gaussian distributions are the Weibull and Gamma distributions. In this work, the approaches developed were implemented on the real-world multi-objective resin bonded sand mould problem. The Pareto frontiers obtained were benchmarked using two metrics; the hyper volume indicator (HVI) and the proposed Average Explorative Rate (AER) metric. Detail discussions from various perspectives on the effects of non-Gaussian random generators in metaheuristics are provided.