Keynote Speaker: Dr Mehmet Emin Aydin
Biography:
Dr Mehmet Emin Aydin is Senior Lecturer in Computer Science at the University of the West of England. He has received BSc, MA and PhD degrees from Istanbul Technical University, Istanbul University and Sakarya University, respectively. His research interests include machine learning, multi agent systems, multi-agent reinforcement learning , metaheuristics, swarm intelligence, resource planning, scheduling and optimization. He has secured several research grants individually and in collaboration, published one edited research book, 120+ journal and conference articles, conducted several guest editorial of special issues and successfully supervised many PhD and post-doctoral supervisions. In addition to being the member of advisory committees of many international conferences, he is editorial board member of various peer-reviewed international journals. He is currently fellow of Higher Education Academy, member of EPSRC College in the UK, and senior member of ACM and IEEE.
Lecture Title:
An explorative journey towards generalisation in problem solving
Lecture Abstract:
Problem solving is one of renown artificial intelligence fields, which has kept attracting research for decades. In the past two decades, metaheuristic optimisation and swarm intelligence algorithms (a.k.a. heuristic search algorithms) have been increasingly popular, particularly in logistic, science, and engineering problems and are recognised as the family of the state-of-art approaches in problem solving. The main challenge appears to be in the speed of algorithmic approximation where many approaches were proposed to accelerate and not to spin in local optima. The balance between explorative and exploitive operations appear to play a crucial role in this process.
The fundamental characteristics of heuristic search algorithms are that they are dependent on parametric structures and search strategies. Some online and offline strategies are employed to obtain optimal configurations for the algorithms. Adaptive operator selection is one of strategies that heuristic algorithms use, which plays a crucial role in the efficiency of heuristic-based problem-solving algorithms, especially, when a pool of operators is used to let algorithms dynamically select operators to produce new candidate solutions. A sequence of selected operators is built up throughout the search which impacts the success of the algorithms. Successive operators in a bespoke sequence can be complementary and therefore diversify the search while randomly selected operators are not expected to behave in this way. However, recent research demonstrates that the state-of-art adaptive selection schemes have been proposed to select the best next operator without considering the problem state in the process. In addition, the inefficiencies in problem solving with search procedures can be avoided using the experiences gained while search is undergoing utilising machine learning approaches. In the field of machine learning, reinforcement learning refers to goal-oriented algorithms, which learn from the environment how to achieve a goal.
Recently, we conducted research on that if we can build up more generic approach which can retain gained experiences and exploit for avoidance of inefficiencies in the search process. Binary problem solving is one of the approaches alongside use of adaptively selected multiple operators. Secondly, reinforcement learning was proposed to be embedded in search algorithms for taking the problem state on board in operator selection process. The proposed approach implies mapping the problem states to the best fitting operators in the pool so as to achieve higher diversity and shape up an optimum operator sequence throughout the search process. However, despite significant performance improvement, it is observed that learned information may be transferred from one problem-solving procedure to another. The third lag of our studies was to investigate how to retain the gained experiences for longer use and for transferrable use across problems with various size and types. The research results in these regard will be summarised over solving few types of combinatorial optimisation problems.
