Research Day

Research Day

On Tuesday June 21, the Flanders AI Research Program will organise its Research Day for all researchers related to the program. The event will offer you sufficient possibilities to share your ideas and opinions on the program and learn from peers through workshops, demos and networking.

Registrations are closed. Too late to register? Please send a message to and we will try to add you to the list manually .




09:30 - 10:00 Welcome by Sabine Demey Auditorium
10:00 - 10:30 Keynote by Prof. Benoît Frénay, UNamur Auditorium
10:30 - 11:30 Posterbreak  
11:30 - 13:00 Presentation on Research challenges 2, 3 and 4 Auditorium
  A gentle introduction to Bayesian optimization Room 2
  Industrial use cases Room 3
  Explaining models and predictions Room 4 - 1st floor
  AI on Time Series Green Room
  Various thematic tables Dining room
13:00 - 14:00 Lunch  
14:00 - 14:30 Keynote "An ethical view on AI", by Prof. Lode Lauwaert, KU Leuven Auditorium
14:45 - 15:45 Posterbreak  
15:45 - 17:00 Presentation on Research challenge 1 Auditorium
  AI & Game Theory Room 2
  ZigZag Room 3
  AI & Logic Room 4 - 1st floor
  Tensors for beginners Green Room
  Various thematic tables Dining room
17:00 - 18:30 Networking reception  



We have more than 80 posters present for you to discover. Because of a limited available space, the posters are shown in 2 random groups. Each poster will have a number. The presenters of the posters with an even number will present their poster during the morning session (10h30 - 11h30); the presenters of the posters with an uneven number, at the afternoon sessions (14h45 - 15h45).



AI on Time Series: introducing the COBRAS/DTAIDistance toolbox - led by Wannes Meert
A practical introduction in time series clustering, unsupervised and semi-supervised. We will build up a Python pipeline using our own toolboxes (DTAIDistance, Cobras) and other toolboxes (e.g., Pandas, PyClustering, sklearn, sktime).

A gentle introduction to Bayesian optimization: TRIESTE - led by Ivo Couckuyt
Bayesian Optimization (BO) is a powerful framework for optimization and active learning when data is expensive to obtain. For example, running high-fidelity simulations in engineering design, or tuning the hyperparameters of a deep learning model can take hours to days. In this tutorial, we give a theoretical introduction to BO and explain basic concepts using easy-to-understand illustrations. Afterward, we present the Trieste library – a state-of-the-art BO library in Tensorflow – by demonstrating it on various examples from engineering. 

Explaining models and predictions - led by Tijl De Bie
Explaining the properties and inner workings of models is useful to ascertain their robustness and generalizability, and thus to create trust in their usefulness to make predictions and support decision-making. This workshop will consist of short presentations (pitches) from the various research groups that develop or apply methods for explaining models and predictions, followed by a roundtable discussion on identification of challenges of common interest and opportunities for new collaborations. Please contact us if you are interested in giving a short pitch about recent contributions or challenges faced regarding this topic.

Industrial Use Cases - led by Abdellatif Bey-Temsamani
In the Flanders AI Research Program, different industrially relevant use cases have been proposed. In 2021, we realized different demonstrations where your AI research has been integrated in our set-ups and made available for dissemination and outreach to companies and other research communities. Videos about our realizations of 2019-2021 can be found on the Flanders Make AI Research demos YouTube playlist. In the poster sessions of this Research Day, these demos are also presented and explained.  In this workshop, we would like to go through one AI research topic, namely Multimodal AI, and explain you, (i) its relevance to industrial applications, (ii) the data generation step (ii) the AI algorithms trained with the generated data, (iii) the integration steps to deploy the AI algorithms in an Autonomous Tractor. This example can inspire the researchers on how the industrial use cases, can be used with multi-disciplinary research teams to make industry relevant demos.

Tensors for beginners - led by Lieven De Lathauwer
An important research trend is the generalization of “linear algebra and its applications” to “multilinear algebra and its applications”. While linear algebra relies on vectors and matrices, multilinear algebra additionally involves higher-order tensors. Multilinear methods are nonlinear, but in a fairly manageable, linear-like fashion. The tensor setting offers several new mathematical tools for engineers and data analysts. The goal of this tutorial is to introduce these new tools to the participants and show how they can be used.

ZIGZAG - led by Axel Nackaerts
In this workshop, you will learn about, and learn to work with, KU Leuven’s hardware-scheduling design space exploration tool ZIGZAG. For a given neural network topology, ZIGZAG allows to derive the energy and latency spend when executing the network on a specific AI processor. Additionally, the design space of different scheduling/mappings can be explored, or the design space of different hardware accelerators.

AI & Game Theory - led by Eladio Montero and Elias Fernández Domingos
In this workshop, you will participate in a behavioural economics experiment which aims to study how we make decisions in scenarios with shared finite resources, such as the future energy market. In this experiment, you will interact with each other through your own computer. You not only can earn some money but you also help us understand human decision-making.

AI & Logic - led by Bart Bogaerts and Maxime Jakubowski
This workshop with a broad scope aims to bring together researchers interested in logic-based aspects of AI. 

  • 15.45-15.50: Set-up & welcome.
  • 15.50-16.00: Neural Probabilistic Logic Programming (Robin Manhaeve)
  • 16.00-16.10: IDP-Z3: a reasoning engine for FO(.) (Pierre Carbonelle & Simon Vandevelde)
  • 16.10-16.20:  Wasserstein Auto-encoded MDPs: Formal Verification of Efficiently Distilled RL Policies with Many-sided Guarantees (Florent Delgrange)
  • 16.20-16.30: Knowledge representation using dependent type theory (Nick Harley)
  • 16.30-16.40:  Using logic to empower the user (Joost Vennekens)
  • 16.40-16.50: Generating knowledge graphs according to SHACL shapes (Anastasia Dimou)
  • 16.50-17.00: Weaknesses of Classical Logic for Knowledge Representation and what to do about it (Marc Denecker)


Thematic tables

Creative Reasoning - led by Geraint Wiggins
How can AI systems be (autonomously or cooperatively) creative?
How can creativity enhance the value of AI systems in general?
What defines “creativity” in a machine, anyway?

AI for Time Series #1 - led by Katrien De Cock
What kind of time series do you encounter in your research? What application are you working on?
What tools/algorithms do you use? What tools do not exist, but you wish they did?
Who would you like to hear in a seminar?

AI for Time Series #2 - led by Lola Botman, Naomi Wamba and Nick Seeuws
Dealing with errors/noise in a signal
Processing time series: sequence of windows, or view it as a dynamical system?
Representing a signal: time, frequency, or even time-frequency?
Dealing with irregularly sampled time series

Decision Support Systems - led by Giulia Rinaldi and Oscar Mauricio Agudelo
Integrating ML in Decision Support System (DSS): Does it always bring an added value?
Bias and DSS: How much does BIAS count in a Decision-Making Algorithm?
DSS and Data Science: How should a DSS support Data Scientists in projects?

Uncertainty - led by Willem Waegeman, Lorin Werthen-Brabants and Alexandre Arnould
Bayesian methods (e.g. Gaussian processes, Bayesian neural networks)
Methods that differentiate between aleatoric and epistemic uncertainty
Experimental design and active learning

Data Quality Handling - led by Antoon Bronselaer and Guy De Tré
The need for efficient generic tools
The lack of standard knowledge for data quality handling
Temporal aspects of data quality

AI4Healthcare - led by Nick Seeuws and Alexandre Arnould
Benchmark datasets: where to find them & how to use them?
Multimodal approaches
Deep learning vs Machine Learning: when is one better than the other?
The peculiarities of healthcare data, dealing with (small) scale and privacy/security
Validating models, going beyond iid evaluation? (Classical evaluation on a test set)



The Research Day will be organised at the Irish College in Leuven (Janseniusstraat 1, 3000 Leuven).

The location is at the heart of Leuven, only a kilometre walking from the Leuven train station.

The largest / nearest car park is the Q-Park Heilig Hart (Naamsestraat 102, 3000 Leuven).

Janseniusstraat 1, 3000 Leuven