Thijs van de Laar | BIASlab

A Message Passing Realization of Expected Free Energy Minimization

Fri, 17 Oct 2025 00:00:00 +0000

FlexLab

Fri, 01 Aug 2025 00:00:00 +0000

FlexLab is an AI innovation lab that creates sustainable and societal impact by applying advanced AI technologies to enable flexible electricity consumption and ensure that this flexibility can be used to resolve congestion in medium- and low-voltage power grids. FlexLab achieves this by providing a secure environment in which startups and SMEs can test and validate their flexible energy solutions together with value-chain partners.

In addition, FlexLab develops expertise in core AI technologies required for flexible electricity consumption and evaluates these technologies through short-cycle innovation trajectories. This approach helps address grid congestion effectively, improve grid reliability and stability, and contribute to the energy transition toward a CO2-neutral Netherlands by 2050.

An initial prototype will be developed to apply Bayesian networking and deep reinforcement learning to behind-the-meter energy control. Bayesian networking increases the reliability of behind-the-meter energy management, and this work is carried out in collaboration with Zympler (formerly known as Simpl.Energy).

EmbodEAI

Sun, 01 Jun 2025 00:00:00 +0000

State-of-the-art (deep) reinforcement learning systems, for all their fantastic achievements, struggle in real-world tasks that are trivial for humans, especially those involving physical interactions. At the same time these systems consume excessive power for training and operation. That is because they are inefficient with their model representations (many parameters) and their data (big data and many trials for training). We see the (robot) body as an enormous computational resource that is poorly understood and largely underappreciated. But how to harness embodiment remains an important open question.

Signal processing by a physical body is extremely cheap and robust, but specialized; the brain is flexible, but more power-hungry. This results in a design trade-off that leads us to the following two multidisciplinary research questions for this project.

Which continuous-learning processing tasks should be delegated primarily to hardware (body) and which primarily to software (brain), and
how should the brain and the body be designed to capitalize on the potential of embodied intelligence?

We study how we can harness Embodiment as a resource in a next generation of EAI systems.

Supervisory team:

Youri van de Burgt, Associate Professor, Microsystems, ME, Promotor
Irene Kuling, Assistant Professor, Robotics, ME, co-promotor
Thijs van de Laar, Assistant Professor, BIASlab, EE, co-promotor

FEP-Lab

Sun, 01 Jun 2025 00:00:00 +0000

The FEPlab (Free Energy Principle Laboratory) is a collaboration between Eindhoven University of Technology (TU/e) and GN Hearing. The mission of the lab is to ameliorate the participation of hearing-impaired people in formal and informal social settings. The lab will focus its research on transferring a leading physics/neuroscience-based theory about computation in the brain, the Free Energy Principle (FEP), to practical use in human-centered agents such as hearing devices and VR technology.

GN Hearing, which is a globally leading hearing aid manufacturer with a strong research team (of about 20 persons) in Eindhoven, and the TU/e have already been collaborating for many years in BIASlab, which is a research team at the Electrical Engineering department at TU/e. This collaboration has produced theoretical foundations for synthetic FEP-based AI agents. FEPlab has been set up in 2022 and is expected to run until mid-2027. During this time, the partners will continue to develop these FEP agents into a technology that is ready for deployment in the professional hearing device industry.

FEPlab focuses on two Sustainable Development Goals: Goal 3, Good Health and Well-being, and Goal 5, Promote sustained, inclusive, and sustainable economic growth, full and productive employment, and decent work for all. Untreated hearing loss in the elderly increases the risk of developing dementia and Alzheimer’s disease (Ralli et al., 2019) as well as emotional and physical problems (Ciorba et al., 2012). Therefore, this research neatly ties into SDG3 Target 1: reducing premature mortality from non-communicable diseases. Moreover, hearing loss negatively impacts work participation (Svinndal et al., 2018). Hence, this research also ties into SDG8 Target 1: achieve higher levels of economic productivity through technology upgrading and innovation.

The lab comprises experts from different fields of expertise such as Audiology, Autonomous Agents & Robotics, Decision Making, and Machine Learning to tackle the complex multidisciplinary challenges at hand. Socially aware AI and explainable AI are especially important in the lab’s research since the technology needs to be aware of the social context in which it is operating and be able to provide justification for its decisions and actions in a manner that is understandable by humans to ensure its safe use.

Ralli, Massimo, et al. “Hearing loss and Alzheimer’s disease: A Review.” The international tinnitus journal 23.2 (2019): 79-85.
Ciorba, Andrea, et al. “The impact of hearing loss on the quality of life of elderly adults.” Clinical interventions in aging 7 (2012): 159.
Svinndal, Elisabeth Vigrestad, et al. “Hearing loss and work participation: a cross-sectional study in Norway.” International journal of audiology 57.9 (2018): 646-656.

The lab is part of the Innovation Center for Artificial Intelligence

Automating model comparison in factor graphs

Sat, 29 Jul 2023 00:00:00 +0000

Realising Synthetic Active Inference Agents, Part II: Variational Message Updates

Wed, 12 Jul 2023 09:00:00 +0200

Realising Synthetic Active Inference Agents, Part I: Epistemic Objectives and Graphical Specification Language

Fri, 16 Jun 2023 09:00:00 +0200

Active Inference and Epistemic Value in Graphical Models

Wed, 06 Apr 2022 09:00:00 +0200

Auto-AR

Fri, 01 Oct 2021 00:00:00 +0000

Automated Situated Design of Augmented Hearing Reality Algorithms

Variational Message Passing and Local Constraint Manipulation in Factor Graphs

Thu, 24 Jun 2021 16:37:31 +0200

Chance-constrained active inference

Thu, 06 May 2021 14:07:00 +0200

Learning Where to Park

Tue, 15 Sep 2020 16:37:31 +0200

Application of the free energy principle to estimation and control

Tue, 22 Oct 2019 14:50:37 +0100

Automated design of Bayesian signal processing algorithms

Wed, 12 Jun 2019 13:30:00 +0200

Simulating Active Inference Processes by Message Passing

Thu, 28 Mar 2019 16:37:31 +0200

A Factor Graph Approach to Automated Design of Bayesian Signal Processing Algorithms

Tue, 01 Jan 2019 14:50:37 +0100

ForneyLab.jl: Fast and flexible automated inference through message passing in Julia

Fri, 05 Oct 2018 08:07:00 +0200

ForneyLab: A Toolbox for Biologically Plausible Free Energy Minimization in Dynamic Neural Models

Sun, 23 Sep 2018 13:42:00 +0200

ForneyLab.jl: a Julia Toolbox for Factor Graph-based Probabilistic Programming

Wed, 08 Aug 2018 13:36:00 +0200

Variational Stabilized Linear Forgetting in State-Space Models

Sat, 30 Sep 2017 13:10:00 +0200

A probabilistic modeling approach to hearing loss compensation

Tue, 11 Jul 2017 13:19:00 +0200

CoHear

Wed, 01 Feb 2017 00:00:00 +0000

Hearing loss is a very serious health condition that has been associated with early dementia and cognitive decline. Still, hearing aid market penetration is quite low, in particular for the large population that is afflicted with ‘mild-to-moderate’ hearing impairment. This is mainly due to two reasons: Stigma (association with old age) and hearing aids (HA) sound quality. The recent commercial introduction of fashionable ‘hearables’ will likely alleviate the stigma issue. Recent advances in machine learning and cloud computing open new avenues for attacking the sound quality issue for hearing aids. In this project, we intend to develop a (crowd-based) collaborative design approach to improving the sound quality issues for the mild-to-moderately hearing-impaired population.

We expect to build a working prototype for collaboratively designed hearing algorithms that can be applied to a new class of ‘smart hearing devices’ with high appeal to the mild-to-moderately hearing-impaired patient. As an additional benefit, we hope that our technology will ease the transition from hearables to professional hearing aid technology for the moderate-to-profound hearing-impaired population.

A Probabilistic Modeling Approach to Hearing Loss Compensation

Thu, 10 Nov 2016 16:37:31 +0200