A5052975576- Breast Cancer Treatment Studies
- Control Systems and Identification
- Fault Detection and Control Systems
- Advanced Control Systems Optimization
- Advances in Oncology and Radiotherapy
- Breast Lesions and Carcinomas
- Advanced Radiotherapy Techniques
- Cancer survivorship and care
- Global Cancer Incidence and Screening
- Model Reduction and Neural Networks
- COVID-19 and healthcare impacts
- Probabilistic and Robust Engineering Design
- Cancer Treatment and Pharmacology
- Online and Blended Learning
- Body Contouring and Surgery
- Bariatric Surgery and Outcomes
- Distributed Control Multi-Agent Systems
- Migration, Health and Trauma
- Medical Imaging Techniques and Applications
- BRCA gene mutations in cancer
- Adaptive Control of Nonlinear Systems
- Experimental Learning in Engineering
- Stability and Control of Uncertain Systems
- Adversarial Robustness in Machine Learning
- Health and Wellbeing Research
Princess Margaret Cancer Centre
2014-2024
University Health Network
2014-2024
University of Stuttgart
2019-2023
Princess Margaret Hospital
2022
University of Toronto
2005-2021
Hôpital d'Hautepierre
2018
Hôpitaux Universitaires de Strasbourg
2018
St. Michael's Hospital
2015
University of Münster
2003
We consider the problem of designing robust state-feedback controllers for discrete-time linear time-invariant systems, based directly on measured data. The proposed design procedures require no model knowledge, but only a single open-loop data trajectory, which may be affected by noise. First, data-driven characterization uncertain class closed-loop matrices under is derived. By considering this parametrization in control framework, we gains with guarantees stability and performance,...
Due to their susceptibility adversarial perturbations, neural networks (NNs) are hardly used in safety-critical applications. One measure of robustness such perturbations the input is Lipschitz constant input-output map defined by an NN. In this work, we propose a framework train multi-layer NNs while at same time encouraging keeping small, thus addressing issue. More specifically, design optimization scheme based on Alternating Direction Method Multipliers that minimizes not only training...
We consider the problem of computing reachable sets directly from noisy data without a given system model. Several reachability algorithms are presented for different types systems generating data. First, an algorithm over-approximated based on matrix zonotopes is proposed linear systems. Constrained introduced to provide less conservative at cost increased computational expenses and utilized incorporate prior knowledge about unknown Then we extend approach polynomial and, under assumption...
Dissipativity properties have proven to be very valuable for systems analysis and controller design. With the rising amount of available data, there has, therefore, been an increasing interest in determining dissipativity from (measured) trajectories directly, while explicit model system remains undisclosed. Most existing approaches data-driven dissipativity, however, guarantee condition only over a finite-time horizon provide weak or no guarantees on robustness presence noise. In this...
Based on the Fundamental Lemma by Willems et al., entire behaviour of a Linear Time-Invariant (LTI) system can be characterised single data sequence as long input is persistently exciting. This an essential result for data-driven analysis and control. In this work, we aim to generalise LTI Parameter-Varying (LPV) systems. behavioural framework LPV systems, prove that one obtain similar Willems'. representation, i.e., embedding, nonlinear allows application systems beyond linear class.
Due to their susceptibility adversarial perturbations, neural networks (NNs) are hardly used in safety-critical applications. One measure of robustness such perturbations the input is Lipschitz constant input-output map defined by an NN. In this letter, we propose a framework train multi-layer NNs while at same time encouraging keeping small, thus addressing issue. More specifically, design optimization scheme based on Alternating Direction Method Multipliers that minimizes not only training...
There exists a vast amount of literature how dissipativity properties can be exploited to design controllers for stability and performance guarantees the closed loop. With rising availability data, there has therefore been an increasing interest in determining from data as means data-driven systems analysis control with rigorous guarantees. Most existing approaches, however, consider that hold only over finite horizon mostly qualitative statements made presence noisy data. In this work, we...
Background: Improved treatments resulting in a rising number of survivors breast cancer (bca) calls for optimization current specialist-based follow-up care. In the present study, we evaluated well bca with respect to their supportive care needs and attitudes toward various providers, varying settings, or mediated by technology (for example, videoconference e-mail). Methods: A cross-sectional paper survey early-stage pT1–2N0 undergoing posttreatment was completed. Descriptive univariable...
Due to their relevance in controller design, we consider the problem of determining $\mathcal{L}^2$-gain, passivity properties and conic relations an input-output system. While, practice, relation is often undisclosed, data tuples can be sampled by performing (numerical) experiments. Hence, present sampling strategies for discrete time continuous linear time-invariant systems iteratively determine shortage cone with minimal radius that confined to. These are based on gradient dynamical...
Smartphones as our permanent companion seem to be an expedient choice for the implementation of e-learning tools in form smartphone apps due their convenience and accessibility. We hence present two recently published that were developed introductory control course. The serve purposes (i) improving learning progress (ii) revision course content between lectures. In this paper, we explain how align with lecture Introduction Automatic Control, put them into context strategy describe pursued...
Fault detection and isolation is an area of engineering dealing with designing on-line protocols for systems that allow one to identify the existence faults, pinpoint their exact location, overcome them. We consider case multi-agent systems, where faults correspond disappearance links in underlying graph, simulating a communication failure between corresponding agents. study which agents controllers are maximal equilibrium-independent passive (MEIP), use known connection steady-states these...
It is well-known that students learn substantially more if a lecture complemented by active inquiry-based activities and problem solving than from only passively listening to lecture. Naturally, this requires the teacher include elements into course such as laboratories, group projects, tutorials interactive e-learning modules. For most effective teaching concept, all components of work are tightly intertwined together towards same goal learning objectives. Therefore, we show in paper how...
In this paper, we present a method to analyze local and global stability in offset-free setpoint tracking using neural network controllers provide ellipsoidal inner approximations of the corresponding region attraction. We consider feedback interconnection linear plant connection with controller an integrator, which allows for desired piecewise constant reference that enters as external input. Exploiting fact activation functions used networks are slope-restricted, derive matrix inequalities...