INTRODUCTION
This course gives students insights into the recent developments in a fast-growing research line: the integration of techniques from machine learning (ML) and optimization (OPT). The lectures cover different aspects of the integration of ML and optimization. Example topics include:
- Machine learning, including deep learning and reinforcement learning, to solve combinatorial optimization problems; black-box optimization; neural combinatorial optimization; and handling uncertainties of prediction models for decision-making
- Using optimization algorithms for the development of transparent ML models, such as learning optimal decision trees
- Computing explanations for ML model via techniques developed for optimization or constraint reasoning systems
Lectures are complemented by lab sessions with hands-on exercises during the course.
The two-day course will be held in Utrecht in person. To participate, students should have basic knowledge on machine learning and some basic knowledge on optimization (i.e., mathematical programming, or (meta-)heuristics).
DATE:
Tuesday 11 and Wednesday 12 June 2024
LOCATION:
SCIENTIFIC DIRECTORS:
Dr Zaharah Bukhsh, TU/e
Dr Jie Gao, TU Delft
Dr Neil Yorke-Smith, TU Delft
Dr Yingqian Zhang, TU/e
PROGRAMME
11 June 2024
09:00 Doors open, coffee and tea
09:30 Introduction
09:45 Brief introduction to techniques of machine learning and optimization, Wouter Kool, ORTEC
10:15 Introduction to Deep reinforcement learning, Zaharah Bukhsh, TU/e
11:15 Coffee break
11:45 Learning to solve routing problems Wouter Kool
12:30 Lunch
14:00 Lab 1: VRP competition, Wouter Kool
15:00 Coffee break
15:30 Decision tree learning, Sicco Verwer, TU Delft
16:15 Lab 2: Learning decision trees, Sicco Verwer
17:15 Wrap-up of first day
12 June 2024
09:00 Doors open, coffee and tea
09:15 Surrogate-based optimization, Laurens Bliek, TU/e
10:15 Coffee break
10:45 Lab 3: SBO, Laurens Bliek
11:45 Application: analytics for a better world, Joaquim Gromicho, UvA
12:30 Lunch
14:00 Modularization of neural networks for cross-problem optimization, Yaoxin Wu, TU/e
15:15 Coffee break
15:45 Application: Data-driven optimization for ride-hailing vehicle guidance, Jie Gao, TU Delft
16:45 Feedback session
17:00 Course wrap-up
REGISTRATION
Registration for this course is now open. Please note that all available hotel rooms have been booked. If you do wish to be considered for an overnight stay in case of a cancellation, or if you have any other questions, please email Renée.
