Direct Measuring of Exploration and Exploitation in Evolutionary Algorithms using Attraction Basin-based Measures

#evolutionary-computation

Evolutionary algorithms mimic natural processes such as selection, crossover, and mutation to solve a wide range of optimization problems. Effective application of these algorithms requires a solid understanding of the various selection, crossover, and mutation operators. However, exploration and exploitation are fundamental processes—and arguably the most critical concepts—for any search algorithm. Despite their importance, these concepts are not yet well understood among many researchers and practitioners. Furthermore, the direct measurement of exploration and exploitation remains an open problem in evolutionary computation.

This talk first introduces the basic components of evolutionary algorithms, highlights common issues and mistakes encountered by inexperienced users, and discusses their diverse applications. In the second part, a novel direct measure of exploration and exploitation based on attraction basins is presented. Attraction basins are regions of a search space in which each region contains a point, called an attractor, toward which neighbouring points tend to evolve. Each search point can therefore be associated with a specific attraction basin. If a newly generated search point belongs to the same attraction basin as its parent, the search process is considered exploitation; otherwise, it is classified as exploration.

In the final part, the talk demonstrates how the newly developed exploration and exploitation measures can be used to analyze and compare different evolutionary algorithms.

• Add Event to Calendar
  iCal
  Google Calendar

• Engineering Auditorium, National University of Singapore
• Singapore, Singapore
• Singapore

---

• Contact Event Host