CEC Tutorial: Advances in Hyper-Heuristics

Tutorial: Advances in Hyper-Heuristics

Abstract

Over the years the benefits of searching an alternative space, namely, the heuristic space, instead of exploring the solution space directly has been shown. This has been achieved by using hyper-heuristics. As the field of hyper-heuristics has evolved since its inception there have been several advances in the field. The tutorial firstly gives an introductory overview of hyper-heuristics and then delves into the advanced topics. A taxonomy of the different levels of generality that can be attained by a hyper-heuristic is firstly presented. The tutorial then examines assessing hyper-heuristic performance in the context of these hyper-heuristics. The tutorial will then explore machine learning and evolutionary algorithm hyper-heuristics. While a lot of the earlier work in the field focused on discrete optimization, recent advancements include solving continuous optimization problems directly using hyper-heuristics. The tutorial will examine how hyper-heuristics can be used to directly solve continuous optimization problems. The tutorial will also highlight the benefits of transfer learning in evolutionary algorithm hyper-heuristics. Explainable hyper-heuristics, that is the use of XAI to better understand the performance of hyper-heuristics will be examined. The tutorial will conclude with a discussion session on future research in the field of evolutionary algorithm hyper-heuristics.

Tutorial Outline

1. Overview of Hyper-Heuristics
This tutorial firstly provides an overview of hyper-heuristics in general and then more specifically the four types of hyper-heuristics, namely, selection construction, selection perturbative, generation constructive and generation perturbative hyper-heuristics are examined.

2. Evolutionary Algorithm Hyper-Heuristics
This part of the tutorial will describe details of the evolutionary algorithms used and applications for each type of hyper-heuristic. A case study will be presented for each type of hyper-heuristic to illustrate how the hyper-heuristic can be applied. The EvoHyp evolutionary algorithm hyper-heuristic library will be used to demonstrate the implementation of evolutionary algorithm hyper-heuristics for each case study.

3. Taxonomy for Generality Levels in hyper-heuristic
Five levels of generality in hyper-heuristics will be described and a case study for each of the levels will be presented.

4. Assessing Hyper-Heuristic Performance
Assessing hyper-heuristic performance in terms of both generality and optimality will be discussed. An assessment measure for assessing generality, namely, standard deviation of differences, will be presented. Multi-objective assessment of hyper-heuristic performance will also be presented.

5. Machine Learning and Hyper-Heuristics
This part of the tutorial will examine the synergistic relationship between machine learning and evolutionary algorithm hyper-heuristics. Firstly, it will examine how machine learning can be used to improve the performance of hyper-heuristics and secondly how hyper-heuristics can be used to improve the performance of machine learning.

6. Evolutionary Algorithm Hyper-Heuristics for Continuous Optimization
Previous work using hyper-heuristics for solving continuous optimization problems have essentially used hyper-heuristics to design the approaches that are applied to the solve the problem. Here we will look at applying hyper-heuristics directly to solving the problem.

7. Transfer learning in Evolutionary Algorithm Hyper-Heuristics
This part of the tutorial will focus on using transfer learning in evolutionary algorithm hyper-heuristics. Case studies for the different types of hyper-heuristics will be examined.

8. Explainable Evolutionary Algorithm Hyper-Heuristics
The tutorial will examine the recent advances in the use of XAI to better understand and improve hyper-heuristic performance.

9. Future research directions and discussion
This will involve an interactive session on future research directions in the field of evolutionary algorithm hyper-heuristics.

Organizer

Nelishia Pillay
University of Pretoria
nelishia.pillay@up.ac.za

Book

Advances in Hyper-Heuristics