Metaheuristics Stream

Talks:

Solving very large vehicle routing problems with the help of machine learning

Can we solve vehicle routing problems with tens of thousands of customers in a reasonable amount of time? Yes, we can!

In this talk, we will explore how using machine learning algorithms to discover the properties of high-quality solutions, and developing highly specialized heuristic operators that use this knowledge can help us to solve large vehicle routing problems more efficiently.

Presenters:

This stream features two talks by leading experts in metaheuristics.

Kenneth Sörensen, 
University of Antwerp

New Metaheuristic Applications Related to Quantum Computing

Recent years have brought the discovery that the Quadratic Unconstrained Binary Optimisation (QUBO) problem can embrace an exceptional variety of important combinatorial optimisation problems found in industry, science and government.

The QUBO model has emerged as a primary focus of quantum and quantum-related computing and is being intensively explored in initiatives by many organisations in both the private and public sectors. Computational experience is being amassed by both the classical and the quantum computing communities that highlights not only the potential of the QUBO model but also its effectiveness as an alternative to traditional modelling and solution methodologies.

We illustrate the applications of these models and report new developments that have produced algorithms for QUBO problems that significantly outperform all other approaches to date. Experimentation shows that these methods, which can be implemented on classical computers, can solve QUBO problems several times larger and up to three orders of magnitude faster than other methods. We also report advances for solving more general models called QUBO-Plus problems, which expand the range of practical applications that can be successfully handled.