A5011904999- Retinal Imaging and Analysis
- Optical Coherence Tomography Applications
- Machine Learning and Algorithms
- Glaucoma and retinal disorders
- Statistical Methods and Inference
- Probabilistic and Robust Engineering Design
- Computational Drug Discovery Methods
- Medical Image Segmentation Techniques
- Receptor Mechanisms and Signaling
- Chemical Synthesis and Analysis
- Sparse and Compressive Sensing Techniques
- Machine Learning and Data Classification
- Cell Image Analysis Techniques
- Retinal Diseases and Treatments
- Pesticide Residue Analysis and Safety
- Bayesian Methods and Mixture Models
- Computational Geometry and Mesh Generation
- Machine Learning in Materials Science
Heidelberg University
2010-2019
Yahoo (Spain)
2010
Boehringer Ingelheim (Germany)
2009
Screening large libraries of chemical compounds against a biological target, typically receptor or an enzyme, is crucial step in the process drug discovery. Virtual screening (VS) can be seen as ranking problem which prefers many actives possible at top ranking. As standard, current Quantitative Structure−Activity Relationship (QSAR) models apply regression methods to predict level activity for each molecule and then sort them establish In this paper, we propose top-k algorithm (StructRank)...
In the present work we develop a predictive QSAR model for blockade of hERG channel. Additionally, this specific end point is used as test scenario to and evaluate several techniques fusing predictions from multiple regression models. inhibition models which are presented here based on combined data set roughly 550 proprietary 110 public domain compounds. Models built using various statistical learning different sets molecular descriptors. Single Support Vector Regression, Gaussian Process,...
Abstract Non-parametric density estimation with shape restrictions has witnessed a great deal of attention recently. We consider the maximum-likelihood problem estimating log-concave from given finite set empirical data and present computational approach to resulting optimization problem. Our targets ability trade-off costs against accuracy in order alleviate curse dimensionality higher dimensions.
With the introduction of spectral-domain optical coherence tomography (OCT), resulting in a significant increase acquisition speed, fast and accurate segmentation 3-D OCT scans has become evermore important. This paper presents novel probabilistic approach, that models appearance retinal layers as well global shape variations layer boundaries. Given an scan, full posterior distribution over segmentations is approximately inferred using variational method enabling efficient inference terms...
A novel computational approach to log-concave density estimation is proposed. Previous approaches utilize the piecewise-affine parametrization of induced by given sample set. The number parameters as well non-smooth subgradient-based convex optimization for determining maximum likelihood estimate cause long runtimes dimensions $d \geq 2$ and large sets. presented based on mildly non-convex smooth approximations objective function \textit{sparse}, adaptive parametrization. Established...