A5101546537- Natural Language Processing Techniques
- Speech and dialogue systems
- Distributed and Parallel Computing Systems
- Scientific Computing and Data Management
- Parallel Computing and Optimization Techniques
- Advanced Neural Network Applications
- Augmented Reality Applications
- Semantic Web and Ontologies
- Topic Modeling
- Particle accelerators and beam dynamics
- Computer Graphics and Visualization Techniques
- Digital Innovation in Industries
- Image and Video Quality Assessment
- 3D Shape Modeling and Analysis
- Public Administration and Political Analysis
- Advanced Data Storage Technologies
- Superconducting Materials and Applications
- Corporate Governance and Management
- 3D Surveying and Cultural Heritage
- Image Processing Techniques and Applications
- Corporate Social Responsibility and Sustainability
- Recommender Systems and Techniques
- Domain Adaptation and Few-Shot Learning
- Research Data Management Practices
- Service-Oriented Architecture and Web Services
Paul Scherrer Institute
2024-2025
Furtwangen University
2014-2024
Fraunhofer Institute for Physical Measurement Techniques
2021-2024
Fraunhofer Institute for Open Communication Systems
2021-2022
Forschungszentrum Jülich
2022
ETH Zurich
2020-2021
RWTH Aachen University
2019
Saarland University
2014-2015
AstraZeneca (Sweden)
2014
Klinikum Saarbrücken
2013
Most compilers have a single core intermediate representation (IR) (e.g., LLVM) sometimes complemented with vaguely defined IR-like data structures. This IR is commonly low-level and close to machine instructions. As result, optimizations relying on domain-specific information are either not possible or require complex analysis recover the missing information. In contrast, multi-level rewriting instantiates hierarchy of dialects (IRs), lowers programs level-by-level, performs code...
We report on experiments in reference resolution using a decision tree approach. started with standard feature set used previous work, which led to moderate results. A closer examination of the performance features for different forms anaphoric expressions showed good results pronouns, proper names, and poor definite noun phrases. then included cheap, language domain independent based minimum edit distance between strings. This yielded significant improvement data sets consisting phrases...
The human factors approach relies on understanding the properties of capability and limitations under various conditions application that knowledge in designing developing safe systems. Following principles MTO (Man Technology Organization) approach, emphasis should be given to way people interact with technical as well organizational A model describing factor influences relation performance shaping their effect manual ultrasonic inspection had been built a part it empirically tested....
The methods employed to document cystoscopic findings in bladder cancer patients lack accuracy and are subject observer variability. We propose a novel endoimaging system an online documentation platform provide post-procedural 3D reconstructions for improved diagnosis, management follow-up. RaVeNNA4pi consortium is comprised of five industrial partners, two university hospitals technical institutes. These grouped into hardware, software clinical partners according their professional...
<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Video streaming content varies in terms of complexity and requires title-specific encoding settings to achieve a certain visual quality. Classic “one-size-fits-all” ladders ignore video-specific characteristics apply the same across all video files. In worst-case scenario, this approach can lead quality impairments, artifacts, or unnecessarily large media A per-title solution has potential...
Integrated gradients is an explainable AI technique that aims to explain the relationship between a model’s predictions in terms of its features. Adapting this point clouds and semantic segmentation models allows class-wise attribution with respect input This better insight into how model reached prediction. Furthermore, it quantitative analysis much each feature contributes To obtain these attributions, baseline high entropy generated interpolated cloud be visualized. These are then run...
We present a development environment for distributed GPU computing targeted multi-GPU systems, as well graphics clusters. Our system is based on CUDA and logically extends its parallel programming model processors to higher levels of parallelism, namely, the PCI bus network interconnects. While extended API mimics full function set current hardware-including concept global memory-on all distribution layers, underlying communication mechanisms are handled transparently application developer....
Mind maps are useful tools for researchers. They can use mind maps, example, to manage their literature or draft research papers. Dozens of exist create instance, FreeMind, MindManager, and XMind. However, researchers need special features such as the possibility accessing bibliographic databases (e.g., BibTeX) directly from within mapping software. Therefore, we developed SciPlore MindMapping, first tool solely focusing on researchers' needs. It offers all one would expect a standard...
The digitization of health records and cross-institutional data sharing is a necessary precondition to improve clinical research patient care. SMITH project unites several university hospitals medical faculties in order provide informatics solutions for integration communication. In this paper, we focus on requirements elicitation management extracting from heterogeneous subsystems based eHealth standards such as HL7 FHIR IHE profiles.
The evaluation of research beyond scientific impact is increasingly required, but has not yet been widely applied. One reason that data necessary for the societal science are often available in sufficient quantities or suitable form. This paper describes first results from a project develops improved documentation to serve impact. Firstly, we refer need do this, and specific challenges assessment this area. Secondly, describe concept documenting achievements with information system...
In the context of High-Field Magnet (HFM) Programme, hosted at CERN, and Swiss Accelerator Research Technology (CHART) Initiative, PSI, Development (MagDev) project aims to contribute Future Circular Collider (FCC) studies through, among others: (i) research on enabling technologies (ii) high-field low temperature (LTS) high superconducting (HTS) magnets. After a first experience with CD1 magnet, cantedcosine-theta (CCT) Nb <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...
With the growing usage of high quality HD and ultra video content, adaptive bitrate streaming constantly increasing demand for bitrates distribution bandwidth, energy consumption related costs grow exponentially in parallel. As such, it is vital to reduce overall online streaming. In this paper we aim investigate, which parameters influence streaming, on client (device) side, as well during encoding. To conduct systematic investigation, have set up a reproducible measurement environment that...
MMAX21 is a versatile, XML-based annotation tool which has already been used in variety of projects. It also the choice ongoing project DIANA-Summ, deals with anaphora resolution and its application to spoken dialog summarization. The uses ICSI Meeting Corpus (Janin et al., 2003), corpus multi-party dialogs contains considerable amount simultaneous speech. features semi-automatically generated segmentation developers tried track flow by inserting segment starts approximately whenever person...
We present an implemented XML data model and a new, simplified query language for multi-level annotated corpora. The new involves automatic conversion of queries into the underlying, more complicated MMAXQL language. It supports sequential hierarchical, but also associative (e.g. coreferential) relations. has been designed with non-expert users in mind.
Automated Machine Learning (Auto-ML) has primarily been used to optimize network hyperparameters or post-processing parameters, while the most critical component for training a high-quality model, dataset, is usually left untouched. In this paper, we introduce novel approach that applies Auto-ML methods process of generating synthetic datasets machine learning models. Our addresses problem requires complex data generator, and developing tuning generator specific scenario time-consuming...