arXiv subjects

Tom Dhaene

Publications and source records attributed to Tom Dhaene.

15 recordsLinked to original sources

arXiv2026

Bayesian Preference Elicitation: Human-In-The-Loop Optimization of An Active Prosthesis

Tuning active prostheses for people with amputation is time-consuming and relies on metrics that may not fully reflect user needs. We introduce a human-in-the-loop optimization (HILO) approach that leverages direct user preferences to personalize a standard four-parameter prosthesis controller efficiently. Our method employs preference-based Multiobjective Bayesian Optimization that uses a state-or-the-art acquisition function especially designed for preference learning, and includes two algorithmic variants: a discrete version (\textit{EUBO-LineCoSpar}), and a continuous version (\textit{BPE4Prost}). Simulation results on benchmark functions and real-application trials demonstrate efficient convergence, robust preference elicitation, and measurable biomechanical improvements, illustrating the potential of preference-driven tuning for user-centered prosthesis control.

cs.RO↗

arXiv2025

Specification-Oriented Automatic Design of Topologically Agnostic Antenna Structure

Design of antennas for modern applications is a challenging task that combines cognition-driven development of topology intertwined with tuning of its parameters using rigorous numerical optimization. However, the process can be streamlined by neglecting the engineering insight in favor of automatic de-termination of structure geometry. In this work, a specification-oriented design of topologically agnostic antenna is considered. The radiator is developed using a bi-stage algorithm that involves min-max classification of randomly-generated topologies followed by local tuning of the promising designs using a trust-region optimization applied to a feature-based representation of the structure frequency response. The automatically generated antenna is characterized by -10 dB bandwidth of over 600 MHz w.r.t. the center frequency of 6.5 GHz and a dual-lobe radiation pattern. The obtained performance figures make the radiator of use for in-door positioning applications. The design method has been favorably compared against the frequency-based trust-region optimization.

math.NA↗

arXiv2025

TR-Based Antenna Design with Forward FD: the Effects of Step Size on the Optimization Performance

Numerical methods are important tools for design of modern antennas. Trust-region (TR) methods coupled with data-efficient surrogates based on finite differentiation (FD) represent a popular class of antenna design algorithms. However, TR performance is subject to FD setup, which is normally determined a priori based on rules-of-thumb. In this work, the effect of FD perturbations on the performance of TR-based design is evaluated on a case study basis concerning a total of 80 optimizations of a planar antenna structure. The obtained results demonstrate that, for the considered radiator, the performance of the final designs obtained using different FD setups may vary by as much as 18 dB (and by over 4 dB on average). At the same time, the a priori perturbations in a range between 1.5% and 3% (w.r.t. the initial design) seem to be suitable for maintaining (relatively) consistent and high-quality results.

math.NA↗

arXiv2025

Miniaturization-Oriented Design of Spline-Parameterized UWB Antenna for In-Door Positioning Applications

Design of ultra-wideband antennas for in-door localization applications is a challenging task that involves development of geometry that ensures appropriate balance between the size and performance. In this work, a topologically-flexible monopole has been generated using a stratified framework which embeds a gradient-based trust-region (TR) optimization algorithm in a meta-loop that gradually increases structure dimensionality. The optimization has been performed using a composite objective function that maintains acceptable size/performance trade-off. The final design features a reflection below -10 dB within the UWB spectrum and a small footprint of only 182 mm2. The considered method has been benchmarked against a standard TR-based routine executed directly on a multi-dimensional representation of the antenna model.

math.NA↗

arXiv2025

Strategies for Feature-Assisted Development of Topology Agnostic Planar Antennas Using Variable-Fidelity Models

Design of antennas for contemporary applications presents a complex challenge that integrates cognitive-driven topology development with the meticulous adjustment of parameters through rigorous numerical optimization. Nevertheless, the process can be streamlined by emphasizing the automatic determination of structure geometry, potentially reducing the reliance on traditional methods that heavily rely on engineering insight in the course of antenna development. In this work, which is an extension of our conference paper [1], a specification-oriented design of topologically agnostic antennas is considered by utilizing two strategies focused on bandwidth-specific design and bandwidth-enhanced optimization. The process is embedded within a variable-fidelity framework, where the low-fidelity optimization involves classification of randomly generated topologies, followed by their local tuning using a trust-region algorithm applied to a feature-based representation of structure response. The final result is then tuned using just a handful of high-fidelity simulations. The strategies under consideration were verified on a case study basis concerning automatic generation of three radiators with varying parameters. Benchmarks of the algorithm against more standard optimization methods, as well as comparisons of the obtained topologies with respect to state-of-the-art antennas from literature have also been considered.

math.NA↗

arXiv2024

Data-Efficient Interactive Multi-Objective Optimization Using ParEGO

Multi-objective optimization is a widely studied problem in diverse fields, such as engineering and finance, that seeks to identify a set of non-dominated solutions that provide optimal trade-offs among competing objectives. However, the computation of the entire Pareto front can become prohibitively expensive, both in terms of computational resources and time, particularly when dealing with a large number of objectives. In practical applications, decision-makers (DMs) will select a single solution of the Pareto front that aligns with their preferences to be implemented; thus, traditional multi-objective algorithms invest a lot of budget sampling solutions that are not interesting for the DM. In this paper, we propose two novel algorithms that employ Gaussian Processes and advanced discretization methods to efficiently locate the most preferred region of the Pareto front in expensive-to-evaluate problems. Our approach involves interacting with the decision-maker to guide the optimization process towards their preferred trade-offs. Our experimental results demonstrate that our proposed algorithms are effective in finding non-dominated solutions that align with the decision-maker's preferences while maintaining computational efficiency.

cs.NE↗

arXiv2022

$\{\text{PF}\}^2$ES: Parallel Feasible Pareto Frontier Entropy Search for Multi-Objective Bayesian Optimization

We present Parallel Feasible Pareto Frontier Entropy Search ($\{\text{PF}\}^2$ES) -- a novel information-theoretic acquisition function for multi-objective Bayesian optimization supporting unknown constraints and batch query. Due to the complexity of characterizing the mutual information between candidate evaluations and (feasible) Pareto frontiers, existing approaches must either employ crude approximations that significantly hamper their performance or rely on expensive inference schemes that substantially increase the optimization's computational overhead. By instead using a variational lower bound, $\{\text{PF}\}^2$ES provides a low-cost and accurate estimate of the mutual information. We benchmark $\{\text{PF}\}^2$ES against other information-theoretic acquisition functions, demonstrating its competitive performance for optimization across synthetic and real-world design problems.

cs.LG↗

arXiv2020

Sensor Fusion using Backward Shortcut Connections for Sleep Apnea Detection in Multi-Modal Data

Sleep apnea is a common respiratory disorder characterized by breathing pauses during the night. Consequences of untreated sleep apnea can be severe. Still, many people remain undiagnosed due to shortages of hospital beds and trained sleep technicians. To assist in the diagnosis process, automated detection methods are being developed. Recent works have demonstrated that deep learning models can extract useful information from raw respiratory data and that such models can be used as a robust sleep apnea detector. However, trained sleep technicians take into account multiple sensor signals when annotating sleep recordings instead of relying on a single respiratory estimate. To improve the predictive performance and reliability of the models, early and late sensor fusion methods are explored in this work. In addition, a novel late sensor fusion method is proposed which uses backward shortcut connections to improve the learning of the first stages of the models. The performance of these fusion methods is analyzed using CNN as well as LSTM deep learning base-models. The results demonstrate a significant and consistent improvement in predictive performance over the single sensor methods and over the other explored sensor fusion methods, by using the proposed sensor fusion method with backward shortcut connections.

cs.LG↗

arXiv2019

Generating an Explainable ECG Beat Space With Variational Auto-Encoders

Electrocardiogram signals are omnipresent in medicine. A vital aspect in the analysis of this data is the identification and classification of heart beat types which is often done through automated algorithms. Advancements in neural networks and deep learning have led to a high classification accuracy. However, the final adoption of these models into clinical practice is limited due to the black-box nature of the methods. In this work, we explore the use of variational auto-encoders based on linear dense networks to learn human interpretable beat embeddings in time-series data. We demonstrate that using this method, an interpretable and explainable ECG beat space can be generated, set up by characteristic base beats.

cs.LG↗

arXiv2017

GPflowOpt: A Bayesian Optimization Library using TensorFlow

A novel Python framework for Bayesian optimization known as GPflowOpt is introduced. The package is based on the popular GPflow library for Gaussian processes, leveraging the benefits of TensorFlow including automatic differentiation, parallelization and GPU computations for Bayesian optimization. Design goals focus on a framework that is easy to extend with custom acquisition functions and models. The framework is thoroughly tested and well documented, and provides scalability. The current released version of GPflowOpt includes some standard single-objective acquisition functions, the state-of-the-art max-value entropy search, as well as a Bayesian multi-objective approach. Finally, it permits easy use of custom modeling strategies implemented in GPflow.

stat.ML↗

arXiv2017

Sensitivity Analysis of Expensive Black-Box Systems Using Metamodeling

Simulations are becoming ever more common as a tool for designing complex products. Sensitivity analysis techniques can be applied to these simulations to gain insight, or to reduce the complexity of the problem at hand. However, these simulators are often expensive to evaluate and sensitivity analysis typically requires a large amount of evaluations. Metamodeling has been successfully applied in the past to reduce the amount of required evaluations for design tasks such as optimization and design space exploration. In this paper, we propose a novel sensitivity analysis algorithm for variance and derivative based indices using sequential sampling and metamodeling. Several stopping criteria are proposed and investigated to keep the total number of evaluations minimal. The results show that both variance and derivative based techniques can be accurately computed with a minimal amount of evaluations using fast metamodels and FLOLA-Voronoi or density sequential sampling algorithms.

cs.OH↗

arXiv2016

Positive blood culture detection in time series data using a BiLSTM network

The presence of bacteria or fungi in the bloodstream of patients is abnormal and can lead to life-threatening conditions. A computational model based on a bidirectional long short-term memory artificial neural network, is explored to assist doctors in the intensive care unit to predict whether examination of blood cultures of patients will return positive. As input it uses nine monitored clinical parameters, presented as time series data, collected from 2177 ICU admissions at the Ghent University Hospital. Our main goal is to determine if general machine learning methods and more specific, temporal models, can be used to create an early detection system. This preliminary research obtains an area of 71.95% under the precision recall curve, proving the potential of temporal neural networks in this context.

cs.LG↗

arXiv2016

Hypervolume-based Multi-objective Bayesian Optimization with Student-t Processes

Student-$t$ processes have recently been proposed as an appealing alternative non-parameteric function prior. They feature enhanced flexibility and predictive variance. In this work the use of Student-$t$ processes are explored for multi-objective Bayesian optimization. In particular, an analytical expression for the hypervolume-based probability of improvement is developed for independent Student-$t$ process priors of the objectives. Its effectiveness is shown on a multi-objective optimization problem which is known to be difficult with traditional Gaussian processes.

stat.ML↗

arXiv2016

Active Learning for Approximation of Expensive Functions with Normal Distributed Output Uncertainty

When approximating a black-box function, sampling with active learning focussing on regions with non-linear responses tends to improve accuracy. We present the FLOLA-Voronoi method introduced previously for deterministic responses, and theoretically derive the impact of output uncertainty. The algorithm automatically puts more emphasis on exploration to provide more information to the models.

cs.LG↗

arXiv2016

Fast Calculation of the Knowledge Gradient for Optimization of Deterministic Engineering Simulations

A novel efficient method for computing the Knowledge-Gradient policy for Continuous Parameters (KGCP) for deterministic optimization is derived. The differences with Expected Improvement (EI), a popular choice for Bayesian optimization of deterministic engineering simulations, are explored. Both policies and the Upper Confidence Bound (UCB) policy are compared on a number of benchmark functions including a problem from structural dynamics. It is empirically shown that KGCP has similar performance as the EI policy for many problems, but has better convergence properties for complex (multi-modal) optimization problems as it emphasizes more on exploration when the model is confident about the shape of optimal regions. In addition, the relationship between Maximum Likelihood Estimation (MLE) and slice sampling for estimation of the hyperparameters of the underlying models, and the complexity of the problem at hand, is studied.

cs.CE↗