ECiP @GECCOConf 2023

July 17, 2023 (hybrid)

Evolutionary Computation in Practice

In the Evolutionary Computation in Practice (ECiP) track, well-known speakers with outstanding reputation in academia and industry present background and insider information on how to establish reliable cooperation with industrial partners. They actually run companies or are involved in cooperation between academia and industry. 
If you attend, you will learn multiple ways to extend EC practice beyond the approaches found in textbooks. Experts in real-world optimization with decades of experience share their approaches to creating successful projects for real-world clients. Some of what they do is based on sound project management principles, and some is specific to our type of optimization projects. If you are working in academia and are interested in managing industrial projects, you will receive valuable hints for your own research projects.
In 2023, ECiP will be a hybrid event. We will do our best to enable opportunities for establishing contacts among participants.

Program (Overview)

Day: July 17th, 2023

Time: 11:00 – 13:00 WEST (GMT+1)

11:00 – 11:30Sunith BandaruEvolutionary optimization + Knowledge discovery: A practical combination for intelligent decision supportUniversity of Skövde, Sweden
11:30 – 12:00Ruxandra StoeanSurrogate-based evolutionary algorithms in the optimization of deep learning for semantic segmentationUniversity of Craiova, Romania
12:00 – 12:30Matthias Groß & Thomas Bartz-BeielsteinIMProvT II – Intelligent measurement methods for the energetic process optimization of drinking water supply and distributionUniversity of Applied Sciences of Cologne, Germany
12:30 – 13:00Jens Uwe Brandt, Noah Pütz & Marc HilbertSideslip angle estimation with transfer learning capabilities: A deep learning approach for motorsportsTOYOTA GAZOO Racing Europe


  • Each presentation will be ~ 25 min with additional ~ 5 min of Q&A each.

Sunith Bandaru:
Evolutionary optimization + Knowledge discovery: A practical combination for intelligent decision support


The fields of EMO (Evolutionary Multi-objective Optimization) and MCDM (Multi-Criteria Decision Making) play complementary roles in finding a practical solution to real-world optimization problems which often involve multiple conflicting objectives. While much of the EMO research focuses on obtaining a good representation of trade-off solutions, MCDM methods are mainly aimed at incorporating decision-maker’s preferences in the objective space to identify a single solution for implementation. Naturally, several methods for visualizing the objective function values have been developed to support the task of decision-making, while the variable space has largely been ignored in this regard. However, it can be argued that in many practical problems, knowledge about the impact of variables on the objective space will better inform the decision-maker about their preferences, leading to higher confidence and quality in decisions. It is possible to obtain such knowledge from EMO solutions by customizing various data mining and machine learning methods to take preferences into account. Such a combination of knowledge discovery with EMO methods, together with established visualization and preference elicitation techniques, can lay the foundation for intelligent decision support. This talk will first introduce knowledge discovery in the context of multi-objective optimization, along with relevant forms of explicit knowledge and available methods to extract them from solution sets. This will be followed by a brief demonstration of Mimer – a recently developed intelligent decision support tool for interactive knowledge discovery and knowledge visualization.

Short CV

Sunith Bandaru (Member, ACM; Senior Member, IEEE) is an Associate Professor with the Division of Intelligent Production Systems in the School of Engineering Science and Chair of the Committee for Research Education in Informatics at University of Skövde, Sweden. He received his Bachelor’s degree (2006) in Mechanical Engineering from Jawaharlal Nehru Technological University, India, and Master’s (2008) as well as Ph.D. (2013) degrees from Indian Institute of Technology (IIT) Kanpur, India. His research interests lie at the intersection of evolutionary computation, machine learning and industrial data analytics, where he has been leading the development of algorithms, methods and software tools for knowledge discovery from simulation-based optimization problems, knowledge-driven optimization and intelligent decision support. He has experience in running multiple industry-collaboration projects with manufacturing companies in Sweden.

Ruxandra Stoean:
Surrogate-based evolutionary algorithms in the optimization of deep learning for semantic segmentation


Evolutionary computation has constantly been a first choice for the optimization of the hyperparameters of models in order to enhance their performance for real-world scenarios. With the advent of deep learning and its extensive number of hyperparameters, evolutionary algorithms found a new large playground. However, since the training of a deep learning model is highly resource-consuming, the repetition of the process in order to evaluate every resulting potential solution is not feasible. Surrogate-based evolutionary optimization has in this sense proven to be an efficient solution, with various machine learning approaches offering a partial model of the original deep learner that could be used for the fitness evaluation of its potential hyperparameter configurations. This talk will present a practical example of the potential of surrogate-based evolutionary optimization for tuning non-architectural deep learning hyperparameters, i.e., corresponding weights in an imbalanced classification real-world task of semantic segmentation of corrosion compounds from microscopy images of archaeological objects.

Short CV

Ruxandra Stoean is an Associate Professor at the University of Craiova, Romania, and Principal Investigator at the Romanian Institute of Science and Technology, Cluj-Napoca. She received BSc and MSc degrees in Computer Science in 2002 and 2003 from the University of Craiova and a PhD in Computer Science from the University Babes-Bolyai, Cluj-Napoca, with the thesis focused on the optimization of support vector machines through evolutionary computation. Her currents research interests involve the development of optimal deep learning models for images and signals, with applications in medicine, engineering and cultural heritage.

Matthias Groß & Thomas Bartz-Beielstein:
IMProvT II – Intelligent measurement methods for the energetic process optimization of drinking water supply and distribution


The goal of the “IMProvT II” project (Intelligent Measurement Methods for Process Optimization of Drinking Water Provision and Distribution) is to develop a digitalization platform for the water industry to optimize the energy efficiency of drinking water provision and distribution. The project aims to address the high variable costs, which account for the largest share of energy costs in the overall expenses of water supply. Rising energy prices have led to a significant increase in water supply fees, creating a burden for consumers. Therefore, ensuring an energy-efficient operation of water supply systems is crucial. To achieve this, a central and flexible data platform is needed to integrate relevant process and management information from water extraction and distribution networks. By consolidating this data, it can be made available to other services for analysis and operational optimization. The platform will utilize methods of Artificial Intelligence (AI) and Machine Learning (ML) to analyze, predict, optimize, and control various aspects of the water supply system. The project also focuses on data privacy and security, especially for Critical Infrastructures (KRITIS), and ensures that the digitalization platform does not directly interfere with the control system. Instead, it acts as a parallel assistant system. Additionally, the project aims to design and analyze business models for digitalization in the water industry to facilitate knowledge transfer between research, water associations, and the industry. The project includes a case study involving two waterworks to demonstrate the potential energy savings and financial benefits. The developed solution can potentially reduce the energy consumption of water provision by 11-14%, resulting in significant cost savings. Furthermore, this reduction in energy demand would lead to a corresponding decrease in CO2 emissions. The project aligns with the maturity model for digitalization in the water industry and focuses on advancing stages 4, 5, and 6, which involve understanding the system, forecasting, and achieving autonomous control through adaptive regulation and assistance systems.

Short CV – Matthias Groß

Matthias Groß is a research assistant at the Institute for Data Science, Engineering, and Analytics (IDE+A) at TH Köln. He obtained his BSc and MSc degrees in Computer Sciences from the same university in 2017 and 2022, respectively. In 2017, he also began assisting in various study courses at the Advanced Media Institute of TH Köln. After teaching the course “Image-Based Computer Graphics” in late 2022, he transitioned to the IDE+A institute in March 2023 to contribute to the IMProvT II project.

Short CV – Thomas Bartz-Beielstein

Prof. Dr. Thomas Bartz-Beielstein is a well-known expert in artificial intelligence with over 30 years of experience. As Professor of Applied Mathematics at TH Köln and Head of the Institute for Data Science, Engineering and Analytics (IDE+A), he specializes in research in artificial intelligence, machine learning, simulation and optimization. He is the developer of Sequential Parameter Optimization (SPO), an innovative approach that integrates surrogate model-based optimization and evolutionary computation. His work covers a wide range of topics in applied mathematics and statistics, design of experiments, simulation-based optimization, and applications in areas such as water management, elevator control, and mechanical engineering.

Jens Uwe Brandt, Noah Pütz & Marc Hilbert:
Sideslip angle estimation with transfer learning capabilities: A deep learning approach for motorsports


Knowing the sideslip angle of a race car at every point in time is a crucial requirement to compete in modern day competitions. Unfortunately, measuring the sideslip angle directly is prohibited under the regulations in endurance racing and too expensive to be deployed in mass-produced cars. Being limited by the available data and capabilities of embedded hardware, how can one get a good estimate of the sideslip angle?
Considering the inherent non-linear behavior of vehicle dynamics, neural network approaches have gained popularity, particularly in high-speed environments. However, these approaches often struggle to generalize effectively across different car setups and tracks. Our research focuses on the transfer of knowledge from one track to others, where labeled data is sparse or even not available. By doing so, the findings could have the potential to enhance performance not only in motorsports, but also in the long term for electric cars, where more complex control strategies are possible.
Attendees will have the opportunity to gain insights into the practical implications of our research and its potential to improve virtual sensing in the literally most fast paced environment.

Short CV – Jens Uwe Brandt & Noah Pütz

Noah Pütz and Jens Brandt are both Master’s students specializing in Automation and IT, with a focus on Data Science, Engineering, and Analytics. They are working on their Master’s thesis in collaboration with Toyota Gazoo Racing and the TH Köln’s “KI Forschungscluster”. With their expertise in AI and deep learning, Noah and Jens are developing advanced models for estimating sideslip angles in motorsports, leveraging the power of transfer learning to adapt to changing conditions.

Noah Pütz has gained work experience at reputable companies such as BMW Motorrad, Siemens Technology, and the University of Applied Sciences Cologne. He has been involved in various projects and research internships focused on topics like hyperparameter tuning, artificial intelligence, and data science.

Jens Brandt has gained valuable work experience at esteemed organizations such as HF Mixing Group, AGCO Corp., and Porsche Consulting. During his tenure at these companies, Jens actively contributed to various projects and initiatives, showcasing his expertise in areas such as automation, engineering, and process optimization.

Short CV – Marc Hilbert

Dr.-Ing. Marc Hilbert is a seasoned professional with a wealth of experience in artificial intelligence (AI), machine learning, and data analytics. He currently holds the position of Senior Manager and AI Strategy Lead at TOYOTA GAZOO Racing Europe in Cologne, Germany. In this role since May 2022, he is responsible for shaping the organization’s AI strategy and driving innovation in the context of motorsports.
In the automotive industry, Marc’s expertise shines through his role as Team Lead for Machine Learning for Engineering and Production at Volkswagen AG in Munich, Germany. From October 2017 to March 2022, he led a team of data science experts, driving innovations in engineering, connected car technologies, and production applications within the Volkswagen Group.
Marc’s career also includes roles as a Project Manager and Data Scientist, where he worked on analytic use cases across various brands and business areas within the Volkswagen Group. He has a strong background in vehicle function development, predictive maintenance, spare part optimization, and logistics.
IN parallel to his work with TOYOTA GAZOO Racing Europe, Marc works as a Lecturer in Machine Learning at Leiden University in the Netherlands. Since November 2018, he has been teaching a master’s course titled “Machine Learning for Business Analytics”. Marc has also made significant contributions to RWTH Aachen University in Germany, starting in April 2011. He currently serves as a Lecturer in Data Analytics, where he focused on teaching students about data analytics and machine learning applications.
With a solid foundation in academia and industry, Marc Hilbert has successfully applied his knowledge and leadership skills to drive AI and machine learning advancements. His contributions have spanned multiple domains, making a significant impact in the fields of motorsports, business analytics, and the automotive industry.


  • Thomas Bartz-Beielstein, IDE+A, TH Köln, Germany
  • Bogdan Filipic, Jozef Stefan Institute, Slovenia
  • Richard Schulz, IDE+A, TH Köln, Germany


GECCO Schedule at a Glance (