*Project Title: Learn2Opt — Learning to Optimize: A Sustainable Approach*
Host Institution: ESSEC Business School
Location: Paris Area, France
Starting Date: September–November 2025 (flexible)
Duration: 18 months (1.5 years), with a possible extension of up to 1
additional year
*Project Summary:*
The proposed research aims to develop innovative, efficient, and
sustainable techniques for optimization enhanced by machine learning (ML),
focusing on decomposition methods such as column generation and Benders
decomposition. These methods are essential for solving complex large-scale
decision-making problems, but their high computational demands often limit
their practical application. A major challenge in current methodological
evaluations is the neglect of the time and effort required to train ML
models. In practice, training these models can require significant
computational resources—often thousands of GPU hours—resulting in high
energy consumption and substantial carbon emissions.
This project seeks to address this gap by integrating Active Learning (AL)
and Reinforcement Learning (RL) to improve the efficiency and effectiveness
of optimization processes. AL will reduce training times by focusing on the
most informative data, while RL will dynamically guide the optimization
process. By explicitly incorporating training effort as a key evaluation
criterion alongside solution quality, the project aspires to establish a
more comprehensive and sustainable standard in the field.
To validate our approach, we will use the Stochastic Dial-a-Ride Problem
(SDARP) as a test case. SDARP involves optimizing vehicle routing under
uncertain, real-time conditions, reflecting the complexities of real-world
applications such as urban transportation systems.
*Position Description:*
We are seeking a highly motivated postdoctoral researcher with expertise
in mathematical optimization and machine learning. The successful candidate
will work on designing and implementing ML-augmented optimization
frameworks, with a particular focus on integrating machine learning models
into decomposition techniques and developing data-efficient training
strategies based on active learning. The position also involves
contributing to the definition of benchmark problems and the evaluation of
novel methodologies on both synthetic and real-world instances.
Main Tasks:
- Develop efficient ML models to approximate subproblems in
decomposition-based algorithms
- Integrate Active Learning (AL) techniques to reduce the training dataset
size
- Implement and benchmark optimization pipelines on real and synthetic
instances of SDARP
- Collaborate with international experts and contribute to scientific
publications
Candidate Profile:
-PhD in Operations Research, Machine Learning, Applied Mathematics,
Computer Science, or a related field
-Experience with mathematical optimization and machine learning
- Knowledge of decomposition methods, active learning, or reinforcement
learning is considered a plus
- Excellent coding skills (e.g., Python, Julia)
-Excellent written and oral communication skills in English
Advisory Team:
- Prof. Emiliano Traversi (ESSEC Business School, France)
- Prof. Morteza Haghir Chehreghani (Chalmers University / University of
Gothenburg, Sweden)
- Prof. Ashkan Panahi (Chalmers University / University of Gothenburg,
Sweden)
How to Apply:
Send your application to emiliano.traversi@essec.edu with the subject
"Postdoc Application - Learn2Opt".
Your application should include:
- A detailed CV
- A short motivation letter (1 page)
- Names and contacts of two references
- (Optional) up to two representative publication
The position will remain open until filled.
ESSEC Business School is an equal opportunity employer and values diversity
in its workforce.
**********************************************************
*
* Contributions to be spread via DMANET are submitted to
*
* DMANET@zpr.uni-koeln.de
*
* Replies to a message carried on DMANET should NOT be
* addressed to DMANET but to the original sender. The
* original sender, however, is invited to prepare an
* update of the replies received and to communicate it
* via DMANET.
*
* DISCRETE MATHEMATICS AND ALGORITHMS NETWORK (DMANET)
* http://www.zaik.uni-koeln.de/AFS/publications/dmanet/
*
**********************************************************
Posts
