Speaker:  Roberto Maria Rosati   - University of Udine, Italy

Thời gian: 20h-21h tối thứ 5, ngày 6/4/2023

Abstract:
Multi-neighborhood search is based on the composition of multiple local search neighborhoods. This provides a better connectivity in the search space and enables the ability to explore different local minima, while also reducing the risk of getting stuck in a particular region of the search space. In this talk, we review the traditional techniques and discuss the most recent advances in multi-neighborhood search, including the application of reinforcement learning in stochastic multi-neighborhood search for the adaptive tuning of operator weights. Finally, we show some concrete applications of multi-neighborhood simulated annealing to timetabling and scheduling problems, where the multi-neighborhood has proven to be a crucial component for the solution of those hard combinatorial problems.

Short bio: 

Roberto Maria Rosati is a PhD fellow at the Intelligent Optimization Lab, University of Udine (Italy), where he is supervised by Prof. Andrea Schaerf. His research focuses on the design of multi-neighborhood and hybrid metaheuristics for real-world scheduling and timetabling problems. He was visiting student at the Artificial Intelligence Research Institute (IIIA-CSIC), in Barcelona, and at the Technical University of Vienna (TU Wien). Prior to his PhD, he worked as information technology consultant and project manager, gaining first-hand experience in bringing scheduling solutions to the industry.