A5040144256- Topic Modeling
- Advanced Graph Neural Networks
- Earthquake Detection and Analysis
- Advanced Image and Video Retrieval Techniques
- Solar and Space Plasma Dynamics
- Ionosphere and magnetosphere dynamics
- Data Quality and Management
- Anomaly Detection Techniques and Applications
- Data Management and Algorithms
- Complex Network Analysis Techniques
- Bayesian Modeling and Causal Inference
- Astrophysics and Cosmic Phenomena
- Satellite Image Processing and Photogrammetry
- Statistical and numerical algorithms
- Hydrology and Watershed Management Studies
- Infrared Target Detection Methodologies
- Multimodal Machine Learning Applications
- Software Engineering Research
- Gamma-ray bursts and supernovae
- Explainable Artificial Intelligence (XAI)
- Semantic Web and Ontologies
- Advanced X-ray and CT Imaging
- Geophysics and Gravity Measurements
- Neural Networks and Applications
- Scientific Computing and Data Management
LMU Klinikum
2020-2023
Ludwig-Maximilians-Universität München
2018-2023
Munich Center for Machine Learning
2023
Data:Lab Munich (Germany)
2023
Institute of Informatics of the Slovak Academy of Sciences
2021
The heterogeneity in recently published knowledge graph embedding models' implementations, training, and evaluation has made fair thorough comparisons difficult. To assess the reproducibility of previously results, we re-implemented evaluated 21 models PyKEEN software package. In this paper, outline which results could be reproduced with their reported hyper-parameters, only alternate not at all, as well provide insight to why might case. We then performed a large-scale benchmarking on four...
Recently, knowledge graph embeddings (KGEs) received significant attention, and several software libraries have been developed for training evaluating KGEs. While each of them addresses specific needs, we re-designed re-implemented PyKEEN, one the first KGE libraries, in a community effort. PyKEEN 1.0 enables users to compose embedding models (KGEMs) based on wide range interaction models, approaches, loss functions, permits explicit modeling inverse relations. Besides, an automatic memory...
Abstract The Earth’s ionosphere affects the propagation of signals from Global Navigation Satellite Systems (GNSS). Due to non-uniform coverage available observations and complicated dynamics region, developing accurate models has been a long-standing challenge. Here, we present Neural network-based model Electron density in Topside (NET), which is constructed using 19 years GNSS radio occultation data. NET tested against situ measurements several missions shows excellent agreement with...
Abstract The radiation belts of the Earth, filled with energetic electrons, comprise complex and dynamic systems that pose a significant threat to satellite operation. While various models electron flux both for low relativistic energies have been developed, behavior medium energy (120–600 keV) especially in MEO region, remains poorly quantified. At these energies, electrons are driven by convective diffusive transport, their prediction usually requires sophisticated 4D modeling codes. In...
Obtaining labels for medical (image) data requires scarce and expensive experts. Moreover, due to ambiguous symptoms, single images rarely suffice correctly diagnose a condition. Instead, it often take additional background information such as the patient's history or test results into account. Hence, instead of focusing on uninterpretable black-box systems delivering an uncertain final diagnosis in end-to-end-fashion, we investigate how unsupervised methods trained without anomalies can be...
Peer reviewing is a central process in modern research and essential for ensuring high quality reliability of published work. At the same time, it time-consuming increasing interest emerging fields often results review workload, especially senior researchers this area. How to cope with problem an open question vividly discussed across all major conferences. In work, we propose Argument Mining based approach assistance editors, meta-reviewers, reviewers. We demonstrate that decision field...
Visual relation detection methods rely on object information extracted from RGB images such as 2D bounding boxes, feature maps, and predicted class probabilities. We argue that depth maps can additionally provide valuable relations, e.g. helping to detect not only spatial standing behind, but also non-spatial holding. In this work, we study the effect of using different features with a focus maps. To enable study, release new synthetic dataset VG-Depth, an extension Genome (VG). note given...
An emerging trend in representation learning over knowledge graphs (KGs) moves beyond transductive link prediction tasks a fixed set of known entities favor inductive that imply training on one graph and performing inference new with unseen entities. In setups, node features are often not available shallow entity embedding matrices is meaningless as they cannot be used at time Despite the growing interest, there enough benchmarks for evaluating methods. this work, we introduce ILPC 2022,...
The link prediction task on knowledge graphs without explicit negative triples in the training data motivates usage of rank-based metrics. Here, we review existing metrics and propose desiderata for improved to address lack interpretability comparability datasets different sizes properties. We introduce a simple theoretical framework upon which investigate two avenues improvements via alternative aggregation functions concepts from probability theory. finally several new that are more easily...
In this work, we take a closer look at the evaluation of two families methods for enriching information from knowledge graphs: Link Prediction and Entity Alignment. current experimental setting, multiple different scores are employed to assess aspects model performance. We analyze informativeness these measures identify several shortcomings. particular, demonstrate that all existing can hardly be used compare results across datasets. Therefore, propose adjustments empirically how supports...
The spatial distribution of energetic protons contributes towards the understanding magnetospheric dynamics. Based upon 17 years Cluster/RAPID observations, we have derived machine learning-based models to predict proton intensities at energies from 28 1,885 keV in 3D terrestrial magnetosphere radial distances between 6 and 22 RE. We used satellite location indices for solar, solar wind geomagnetic activity as predictors. results demonstrate that neural network (multi-layer perceptron...
Multi-hop logical reasoning is an established problem in the field of representation learning on knowledge graphs (KGs). It subsumes both one-hop link prediction as well other more complex types queries. Existing algorithms operate only classical, triple-based graphs, whereas modern KGs often employ a hyper-relational modeling paradigm. In this paradigm, typed edges may have several key-value pairs known qualifiers that provide fine-grained context for facts. queries, modifies meaning...
Given a very large multimedia database, how to process k-nearest-neighbor queries efficiently? While the sequential scan is one of most obvious solutions for small-to-moderate databases, it becomes practically infeasible when database size grows. Concomitant with volume and velocity data, databases are frequently endowed complex distance-based similarity model that supports content-based data access in an adjustable adaptive manner. Typical many state-of-the-art models at least quadratic...
Abstract One of the major and unfortunately unforeseen sources background for current generation X-ray telescopes are few tens to hundreds keV (soft) protons concentrated by mirrors. such telescope is European Space Agency’s (ESA) Multi-Mirror Mission (XMM-Newton). Its observing time lost due contamination about 40%. This loss affects all broad science goals this observatory, ranging from cosmology astrophysics neutron stars black holes. The soft-proton could dramatically impact future large...