A5032940986- Machine Learning and Data Classification
- Data Analysis with R
- Advanced Multi-Objective Optimization Algorithms
- Metaheuristic Optimization Algorithms Research
- Neural Networks and Applications
- Statistical Methods and Inference
- Data Mining Algorithms and Applications
- Parallel Computing and Optimization Techniques
- Simulation Techniques and Applications
- Computability, Logic, AI Algorithms
- Statistical Methods in Clinical Trials
- Advanced Statistical Methods and Models
- Evolutionary Algorithms and Applications
- Advanced Clustering Algorithms Research
- Coronary Interventions and Diagnostics
- Radiomics and Machine Learning in Medical Imaging
- Gene expression and cancer classification
- Reproductive Biology and Fertility
- Advanced MRI Techniques and Applications
- Cardiac Imaging and Diagnostics
- Optimal Experimental Design Methods
- Sperm and Testicular Function
- Caveolin-1 and cellular processes
- Imbalanced Data Classification Techniques
- Topological and Geometric Data Analysis
Stanford University
2024-2025
Technical University of Munich
2025
Stanford Medicine
2023-2024
TU Dortmund University
2005-2023
Ludwig-Maximilians-Universität München
2016-2023
University of Veterinary Medicine Hannover, Foundation
1988-2016
Institute of Pathology Celle
2014
Abstract Most machine learning algorithms are configured by a set of hyperparameters whose values must be carefully chosen and which often considerably impact performance. To avoid time‐consuming irreproducible manual process trial‐and‐error to find well‐performing hyperparameter configurations, various automatic optimization (HPO) methods—for example, based on resampling error estimation for supervised learning—can employed. After introducing HPO from general perspective, this paper reviews...
We present mlrMBO, a flexible and comprehensive R toolbox for model-based optimization (MBO), also known as Bayesian optimization, which addresses the problem of expensive black-box by approximating given objective function through surrogate regression model. It is designed both single- multi-objective with mixed continuous, categorical conditional parameters. Additional features include multi-point batch proposal, parallelization, visualization, logging error-handling. mlrMBO implemented in...
Hyperparameter optimization constitutes a large part of typical modern machine learning (ML) workflows. This arises from the fact that ML methods and corresponding preprocessing steps often only yield optimal performance when hyperparameters are properly tuned. But in many applications, we not interested optimizing pipelines solely for predictive accuracy; additional metrics or constraints must be considered determining an configuration, resulting multi-objective problem. is neglected...
Bayesian boundary condition (BC) calibration approaches from clinical measurements have successfully quantified inherent uncertainties in cardiovascular fluid dynamics simulations. However, estimating the posterior distribution for all BC parameters three-dimensional (3D) simulations has been unattainable due to infeasible computational demand. We propose an efficient method identify Windkessel parameter posteriors: only evaluate 3D model once initial choice of BCs and use result create a...
Most machine learning algorithms are configured by one or several hyperparameters that must be carefully chosen and often considerably impact performance. To avoid a time consuming unreproducible manual trial-and-error process to find well-performing hyperparameter configurations, various automatic optimization (HPO) methods, e.g., based on resampling error estimation for supervised learning, can employed. After introducing HPO from general perspective, this paper reviews important methods...
Abstract The substantial computational cost of high‐fidelity models in numerical hemodynamics has, so far, relegated their use mainly to offline treatment planning. New breakthroughs data‐driven architectures and optimization techniques for fast surrogate modeling provide an exciting opportunity overcome these limitations, enabling the such technology time‐critical decisions. We discuss application repair multiple stenosis peripheral pulmonary artery disease through either transcatheter...
Computational simulations of coronary artery blood flow, using anatomical models based on clinical imaging, are an emerging non-invasive tool for personalized treatment planning. However, current contend with two related challenges - incomplete anatomies in image-based due to the exclusion arteries smaller than imaging resolution, and lack flow distributions informed by patient-specific imaging. We introduce a data-enabled, multi-scale simulation framework spanning large myocardial...
Hyperparameter optimization constitutes a large part of typical modern machine learning workflows. This arises from the fact that methods and corresponding preprocessing steps often only yield optimal performance when hyperparameters are properly tuned. But in many applications, we not interested optimizing ML pipelines solely for predictive accuracy; additional metrics or constraints must be considered determining an configuration, resulting multi-objective problem. is neglected practice,...
Computational simulations of coronary artery blood flow, using anatomical models based on clinical imaging, are an emerging non-invasive tool for personalized treatment planning. However, current contend with two related challenges - incomplete anatomies in image-based due to the exclusion arteries smaller than imaging resolution, and lack flow distributions informed by patient-specific imaging. We introduce a data-enabled, multi-scale simulation framework spanning large myocardial...
Abstract We propose to use Bayesian optimization (BO) improve the efficiency of design selection process in clinical trials. BO is a method optimize expensive black‐box functions, by using regression as surrogate guide search. In trials, planning test procedures and sample sizes crucial task. A common goal maximize power, given set treatments, corresponding effect sizes, total number samples. From wide range possible designs, we aim select best one short time allow quick decisions. The...
Abstract Motivation To obtain a reliable prediction model for specific cancer subgroup or cohort is often difficult due to limited sample size and, in survival analysis, potentially high censoring rates. Sometimes similar data from other patient subgroups are available, e.g. clinical centers. Simple pooling of all can decrease the variance predicted parameters models, but also increase bias heterogeneity between cohorts. A promising compromise identify those with relationship covariates and...
Abstract The predictive performance of a machine learning model highly depends on the corresponding hyper-parameter setting. Hence, tuning is often indispensable. Normally such requires dedicated to be trained and evaluated centralized data obtain estimate. However, in distributed scenario, it not always possible collect all from nodes due privacy concerns or storage limitations. Moreover, if has transferred through low bandwidth connections reduces time available for tuning. Model-Based...
The big crux with drug delivery to human lungs is that the delivered dose at local site of action unpredictable and very difficult measure, even a posteriori. It highly subject-specific as it depends on lung morphology, disease, breathing, aerosol characteristics. Given these challenges, computational approaches have shown potential, but so far failed due fundamental methodical limitations. We present validate novel in silico model enables prediction deposition throughout entire lung. Its...
Zusammenfassung An 3 Schafen (60–70 kg Lebendmasse) mit Umleitungskanülen im proximalen Colon wurden insgesamt 76 3‐stündige Perfusionsversuche durchgeführt, um die Nettoresorption von Calcium und anorganischem Phosphat aus diesem Darmabschnitt zu bestimmen. Die Perfusionslösungen entsprachen in ihrer Zusammensetzung der Flüssigkeitsphase des Coloninhaltes, wiesen aber für einen Konzentrationsbereich 0–6,56 mmol. l −1 , anorganisches 0–6,53 auf. Für Ca wurde bei Konzentrationen unter 1,20...