Euraxess code:
https://www.euraxess.cz/jobs/415421
The project addresses difficult scheduling and packing problems in the sense of computational complexity, for which classical exact approaches are often impractical at realistic scales. The goal is to design new algorithmic frameworks that combine established tools from Operations Research with modern Machine Learning methods to produce high-quality solutions within acceptable computational times. While rooted in OR, the project requires and will further develop strong competencies in Machine Learning.
Over the past decade, learning-augmented optimization has emerged as a promising paradigm. In exact methods, machine learning has been used to tune solver parameters or guide search in tree-based algorithms for mathematical programming. In heuristic optimization, learning has supported diversification strategies, automated selection of algorithms for specific instances, and even direct construction of solutions. More recently, reinforcement learning and deep learning techniques have been used to guide local search procedures, particularly for transportation and routing problems, demonstrating substantial performance improvements.
This project will focus on scheduling and packing settings, developing general methodologies rather than problem-specific tricks. Two complementary research directions will be pursued.
First, machine learning will be used to improve the parameterization of heuristic algorithms.
Second, learning methods—especially reinforcement learning—will be employed to guide the exploration of solution spaces. This includes selecting promising neighborhoods, prioritizing moves in local search, or constructing solutions incrementally.
The research will build on several successful applications of ML in combinatorial optimization. These include reinforcement learning–guided greedy procedures for graph optimization problems, predictive models for deciding when decomposition techniques should be applied, and classifiers that identify structural characteristics of high-quality solutions. Additional directions involve predicting optimal objective values for complex engineering design problems and developing reinforcement learning–enhanced metaheuristics, such as iterated local search for makespan minimization in advanced manufacturing scheduling.
The developed methods will be evaluated on challenging NP-hard scheduling and packing problems, including both well-studied benchmark problems with strong existing heuristics and more applied, real-world problems where current methods remain insufficient. The objective is to demonstrate that the integration of ML and OR techniques can yield robust improvements across different problem types.
About the group: The Optimization Group, led by Zdenek Hanzalek, focuses on scheduling and combinatorial optimization. The group collaborates strongly with high-tech companies (CEZ – Czech Energy Group, Porsche Engineering Services, EATON, Skoda Auto, ST Microelectronics, Volkswagen, DHL, …). Zdenek is the principal investigator of the Roboprox project and organizer of SchedulingSeminar.com.
[1] Bengio, Y., Lodi, A. Prouvost, A. (2021). Machine Learning for Combinatorial Optimization: a
Methodological Tour d’Horizon, European Journal of Operational Research, 290:405-421.
[2] Nina Mazyavkina, Sergey Sviridov, Sergei Ivanov, Evgeny Burnaev, Reinforcement learning for combinatorial optimization: A survey, Computers & Operations Research, Volume 134, 2021.
[3] Heinz, V.; Hanzálek, Z.; Vilím, P.: Reinforcement Learning for Search Tree Size Minimization in Constraint Programming: New Results on Scheduling Benchmarks, Computers & Industrial Engineering, Volume 209, November 2025, 111413.
[4] Roman Václavík, Antonín Novák, Přemysl Šůcha, Zdeněk Hanzálek, Accelerating the Branch-and-Price Algorithm Using Machine Learning, European Journal of Operational Research, Volume 271, Issue 3, 2018, Pages 1055-1069.
[5] Grus, J.; Hanzalek, Z.: Automated placement of analog integrated circuits using priority-based constructive heuristic, Computers & Operations Research, Volume 167, 106643, July 2024.
Interested candidates are invited to submit their applications at:
https://forms.gle/hj7bMTuKM4ghzPC17 [using Postdoc Position ID: 04-Postdoc-Hanzalek]
The application package should contain: