PhD Scholarship in Machine Learning for Evaluating Constraints in Optimization Algorithms

This project develops state-of-the-art Combinatorial Optimization (CO) algorithms using machine learning techniques and meta-heuristics (e.g., evolutionary algorithms) to learn the values of constraints. Combinatorial Optimization Problems (COPs) are ubiquitous, and many optimization methods have been developed for tackling COPs. In the big data era, we are facing an increasing need to tackle large-scale COPs. Gaining a solid understanding of the usefulness of constraints and making the best use of their values can play a crucial role in tackling this challenge. The project invites applications for a fully funded PhD position focused on developing state-of-the-art optimization algorithms using machine learning techniques to learn the values of constraints. More specifically, the research will investigate how to best characterize constraints in the context of population-based meta-heuristics (e.g., evolutionary algorithms) and use machine learning to learn the values of the constraints for a specific COP, e.g., vehicle routing problem. The objective is to use this knowledge extracted from machine learning to boost the performance of the optimization algorithm. This project will be supported by an ARC Discovery Grant (DP250103251) "Learning to Value Constraints", building on strong publication records in machine learning for combinatorial optimization for the past few years.