A5009516958- Advanced Clustering Algorithms Research
- Data Management and Algorithms
- Face and Expression Recognition
- Complex Network Analysis Techniques
- Anomaly Detection Techniques and Applications
- Time Series Analysis and Forecasting
- Advanced Graph Neural Networks
- Data Mining Algorithms and Applications
- Energy Load and Power Forecasting
- Image Retrieval and Classification Techniques
- Advanced Neuroimaging Techniques and Applications
- Complex Systems and Time Series Analysis
- Functional Brain Connectivity Studies
- Blind Source Separation Techniques
- Bayesian Methods and Mixture Models
- Advanced MRI Techniques and Applications
- Neural Networks and Applications
- Graph Theory and Algorithms
- Algorithms and Data Compression
- Advanced Image and Video Retrieval Techniques
- Bioinformatics and Genomic Networks
- Gene expression and cancer classification
- Dementia and Cognitive Impairment Research
- Parallel Computing and Optimization Techniques
- Domain Adaptation and Few-Shot Learning
University of Vienna
2014-2024
Beilstein-Institut
2023
Technical University of Munich
2008-2021
Ludwig-Maximilians-Universität München
2007-2021
Ocean University of China
2021
FH Campus Wien
2016
Helmholtz Zentrum München
2013-2016
Florida State University
2010-2014
Klinikum rechts der Isar
2010-2012
University of Edinburgh
2011
Subjects with mild cognitive impairment (MCI) have an increased risk to develop Alzheimer's disease (AD). Voxel-based MRI studies demonstrated that widely distributed cortical and subcortical brain areas show atrophic changes in MCI, preceding the onset of AD-type dementia. Here we developed a novel data mining framework combination three different classifiers including support vector machine (SVM), Bayes statistics, voting feature intervals (VFI) derive quantitative index pattern matching...
The perception of pain is characterized by its tremendous intra- and interindividual variability. Different individuals perceive the very same painful event largely differently. Here, we aimed to predict individual sensitivity from brain activity. We repeatedly applied identical stimuli healthy human subjects recorded activity using electroencephalography (EEG). a multivariate pattern analysis time-frequency transformed single-trial EEG responses. Our results show that classifier trained on...
Diffusion tensor imaging (DTI) based assessment of white matter fiber tract integrity can support the diagnosis Alzheimer's disease (AD). The use DTI as a biomarker, however, depends on its applicability in multicenter setting accounting for effects different MRI scanners. We applied multivariate machine learning (ML) to large sample from recently created framework European study Dementia (EDSD). hypothesized that ML approaches may amend acquisition. included 137 patients with clinically...
During the last few years, GPUs have evolved from simple devices for display signal preparation into powerful coprocessors that do not only support typical computer graphics tasks but can also be used general numeric and symbolic computation tasks. As major advantage provide extremely high parallelism combined with a bandwidth in memory transfer at low cost. We want to exploit these dvantages density-based clustering, an important paradigm clustering since algorithms of this category are...
Synchronization is a powerful basic concept in nature regulating large variety of complex processes ranging from the metabolism cell to social behavior groups individuals. Therefore, synchronization phenomena have been extensively studied and models robustly capturing dynamical process proposed, e.g. Extensive Kuramoto Model. Inspired by synchronization, we propose Sync, novel approach clustering. The idea view each data object as phase oscillator simulate interaction objects over time. As...
Alzheimer's disease (AD) progressively degrades the brain's gray and white matter. Changes in matter reflect changes structural connectivity pattern. Here, we established individual networks (ISCNs) to distinguish predementia dementia AD from healthy aging scans. Diffusion tractography was used construct ISCNs with a fully automated procedure for 21 control subjects (HC), 23 patients mild cognitive impairment conversion within 3 years (AD-MCI), 17 dementia. Three typical pattern classifiers...
Can we find heterogeneous clusters hidden in data sets with 80% noise? Although such settings occur the real-world, struggle to methods from abundance of clustering techniques that perform well noise at this level. Indeed, perhaps is enough a departure classical warrant its study as separate problem. In paper present SkinnyDip which, based on Hartigan's elegant dip test unimodality, represents an intriguing approach attractive set properties. Specifically, highly noise-robust, practically...
Graph convolutional network (GCN) with the powerful capacity to explore graph-structural data has gained noticeable success in recent years. Nonetheless, most of existing GCN-based models suffer from notorious over-smoothing issue, owing which shallow networks are extensively adopted. This may be problematic for complex graph datasets because a deeper GCN should beneficial propagating information across remote neighbors. Recent works have devoted effort addressing problems, including...
In current medical practice, patients undergoing depression treatment must wait four to six weeks before a clinician can assess medication response due the delayed noticeable effects of antidepressants. Identification at any earlier stage is great importance, since it reduce emotional and economic burden connected with treatment. We approach prediction patient as classification problem, by utilizing dynamic properties EEG recordings on 7th day present novel framework that applies motif...
This article examines a project dedicated to comprehensively addressing the social impact of digital transformation. The also emphasizes effects transformation on marginalized populations, especially forced migrants and individuals with special needs. involves development open-access course materials titled “Digital Life 1–2–3-4,” which are shared as resources through four MOOCs iMooX platform. primary goal is increase awareness in daily life, such algorithmic bias, inaccessibility, robots...
Hierarchical clustering is a powerful tool for exploratory data analysis, organizing into tree of clusterings from which partition can be chosen. This paper generalizes these ideas by proving that, any reasonable hierarchy, one optimally solve center-based objective over it (such as $k$-means). Moreover, solutions found exceedingly quickly and are themselves necessarily hierarchical. Thus, given cluster tree, we show that access plethora new, equally meaningful hierarchies. Just in standard...
Efficiently processing continuous k-nearest neighbor queries on data streams is important in many application domains, e. g. for network intrusion detection. Usually not all valid objects from the stream can be kept main memory. Therefore, most existing solutions are approximative. In this paper, we propose an efficient method exact k-NN monitoring. Our based three ideas, (1) selecting exactly those which able to become nearest of one or more and storing them a skyline structure, (2)...
Synchronization is a powerful and inherently hierarchical concept regulating large variety of complex processes ranging from the metabolism in cell to opinion formation group individuals. phenomena nature have been widely investigated models concisely describing dynamical synchronization process proposed, e.g., well-known Extensive Kuramoto Model. We explore potential Model for data clustering. regard each object as phase oscillator simulate behavior objects over time. By interaction with...
In high-dimensional feature spaces traditional clustering algorithms tend to break down in terms of efficiency and quality. Nevertheless, the data sets often contain clusters which are hidden various subspaces original space. this paper, we present a selection technique called SURFING (subspaces relevant for clustering) that finds all interesting sorts them by relevance. The sorting is based on quality criterion interestingness subspace using k-nearest neighbor distances objects. As our...
How can we efficiently find a clustering, i.e. concise description of the cluster structure, given data set which contains an unknown number clusters different shape and distribution is contaminated by noise? Most existing clustering methods are restricted to Gaussian model very sensitive noise. If content follows non-Gaussian and/or few outliers belonging no cluster, then computed does not match well true distribution, or unnaturally high required represent set. In this paper propose OCI...
Abstract. The number of wind farms and amount power production in Europe, both on- offshore, have increased rapidly the past years. To ensure grid stability on-time (re)scheduling maintenance tasks to mitigate fees energy trading, accurate predictions speed are needed. Particularly, extreme events high importance farm operators as timely knowledge these can prevent damages offer economic preparedness. This work explores possibility adapting a deep convolutional recurrent neural network...
How do we find a natural clustering of real world point set, which contains an unknown number clusters with different shapes, and may be contaminated by noise? Most algorithms were designed certain assumptions (Gaussianity), they often require the user to give input parameters, are sensitive noise. In this paper, propose robust framework for determining given data based on minimum description length (MDL) principle. The proposed framework, Robust Information-theoretic Clustering (RIC), is...
In situations where class labels are known for a part of the objects, cluster analysis respecting this information, i.e. semi-supervised clustering, can give insight into and structure data set. Several clustering algorithms such as HMRF-K-Means [4], COP-K-Means [26] CCL-algorithm [18] have recently been proposed. Most them extend well-known methods (K-Means [22], Complete Link [17] by enforcing two types constraints: must-links between objects same cannot-links different classes. paper, we...
The ability to deal with uncertain information is becoming increasingly important for modern database applications. Whereas a conventional (certain) object usually represented by vector from multidimensional feature space, an multivariate probability density function (PDF). This PDF can be defined either discretely (e.g. histogram) or continuously in parametric form Gaussian Mixture Model). For of objects, the users expect similar data analysis techniques as certain objects. An technique...
The integrative mining of heterogeneous data and the interpretability result are two most important challenges today's mining. It is commonly agreed in community that, particularly research area clustering, both have not yet received due attention. Only few approaches for clustering objects with mixed-type attributes exist those do consider cluster-specific dependencies between numerical categorical attributes. Likewise, only a papers address problem interpretability: to explain why certain...