image not found

image not found
Lancaster University
Management School
Dept of Management Science
Lancaster, LA1 4YX, UK

Room: B61
T: +44 (0)1524 593117
a.kheiri [at]

I am/was teaching the following modules:

International Lecturing Project Supervision

Title: Hyper/Meta-heuristics for Combinatorial Optimisation Problems

Description: The project has two main parts. The first part is to model an optimisation problem. The difficulty level can be adjusted depending on your background. The problem under consideration can include (but are not limited to) routing, cutting, packing, placement, graph theoretical, timetabling and scheduling problems. The second part would be to investigate and experiment with ways in which the problem could be solved. Meta-heuristics, such as genetic algorithms, simulated annealing and tabu search, are now an established tool for solving hard optimisation problems. A more recent concept is that of "hyper-heuristics", which are algorithms that seek to automate the process of selecting and/or generating (meta-)heuristics. Whereas meta-heuristics draw on Operational Research and Artificial Intelligence, hyper-heuristics draw on Machine Learning and Data Science. A thorough experimentation over a set of problem instances and analysis of results are expected.

Solving/analysing multi/many-objective optimisation problems, developing visualisation tools (e.g., for educational purposes) and explainable models, implementing decision support systems, employing single/population-based selective/generative heuristic, exact or matheuristic approaches (e.g., noising methods), designing techniques for parameters tuning, performing theoretical analysis of heuristics or fitness landscape analysis or applying frequency fitness assignment, generating benchmark instances, considering robustness/fairness of solutions, measuring exploration and exploitation of algorithms, constructing prediction models for algorithm selection using problem-independent/dependant features and/or solution features, considering stochastic and/or dynamic nature of the problem (e.g., the problem instance, constraints and/or the objectives may change over time - note that the change may be gradual or abrupt), using methods for problem reduction (e.g., ML) to reduce the size of an original problem by (for example) predicting whether decision variables belong to an optimal solution or not, assessing energy efficiency of the developed tools, building socially-aware algorithms (e.g., SIQ), developing offline heuristics, implementing multi-threaded heuristics, employing computationally cheap surrogate models to estimate the objective functions, allowing humans to interact to bias the direction of the heuristic algorithms, generating flexible solutions that can adapt to changes, incorporating machine learning in various aspects (e.g., generating initial solutions, selection processes, acceptance criteria, tuning parameters, enhancing exact techniques), creating a tool similar to HyFlex, writing survey papers, etc, can also be considered.

Applicants for this topic should have reasonable mathematical ability, a general interest in optimisation, and strong programming skills. Therefore, the project would suit students who have felt comfortable with the more quantitative/programming courses of our undergraduate/MSc programmes, and would require developing meta- and/or hyper-heuristics for combinatorial optimisation problems. You will spend some time searching, reading and summarising an up-to-date literature on the topic. The projects would require some computer programming, which can be done in the language of your choice but preferably a language appropriate for scientific computing such as Python. Experience with LaTeX/git is also desirable, but is not essential as training can be given.

I have supervised several undergraduate and postgraduate projects in a variety of topics, all of which have had good outcomes. Some of the projects I have supervised have won the department prize for the most outstanding project of the year, some have led to publications and some are client-based projects. You can view previous projects that I have offered in the past for inspiration here.

For PhD applicants, if you are a self-funded student, I have a couple of supervision slots available. Please refer to my personal website to find information about my research interests.