A5004309436- Pulsars and Gravitational Waves Research
- Gamma-ray bursts and supernovae
- Astrophysical Phenomena and Observations
- Advanced Neural Network Applications
- Model Reduction and Neural Networks
- Quantum Chromodynamics and Particle Interactions
- High-pressure geophysics and materials
- Advanced Image and Video Retrieval Techniques
- Visual Attention and Saliency Detection
- Cosmology and Gravitation Theories
- Domain Adaptation and Few-Shot Learning
- Seismology and Earthquake Studies
- Advanced Image Fusion Techniques
- Atomic and Subatomic Physics Research
- Geophysics and Gravity Measurements
- Reservoir Engineering and Simulation Methods
- Stellar, planetary, and galactic studies
- Particle physics theoretical and experimental studies
- Computational Physics and Python Applications
- Advanced Electron Microscopy Techniques and Applications
- Calibration and Measurement Techniques
- Distributed and Parallel Computing Systems
- Optical measurement and interference techniques
- Innovation and Knowledge Management
- Engineering and Materials Science Studies
Max Planck Institute for Intelligent Systems
2021-2025
Max Planck Institute for Gravitational Physics
2022-2024
University of Bonn
2018-2021
Robert Bosch (Australia)
2021
Robert Bosch (Germany)
2019
Ruhr University Bochum
2019
We combine amortized neural posterior estimation with importance sampling for fast and accurate gravitational-wave inference. first generate a rapid proposal the Bayesian using networks, then attach weights based on underlying likelihood prior. This provides (1) corrected free from network inaccuracies, (2) performance diagnostic (the sample efficiency) assessing identifying failure cases, (3) an unbiased estimate of evidence. By establishing this independent verification correction...
Mergers of binary neutron stars emit signals in both the gravitational-wave (GW) and electromagnetic spectra. Famously, 2017 multi-messenger observation GW170817 (refs. 1,2) led to scientific discoveries across cosmology3, nuclear physics4-6 gravity7. Central these results were sky localization distance obtained from GW data, which, case GW170817, helped identify associated transient, AT 2017gfo (ref. 8), 11 h after signal. Fast analysis data is critical for directing time-sensitive...
Deep learning techniques for gravitational-wave parameter estimation have emerged as a fast alternative to standard samplers $\unicode{x2013}$ producing results of comparable accuracy. These approaches (e.g., DINGO) enable amortized inference by training normalizing flow represent the Bayesian posterior conditional on observed data. By conditioning also noise power spectral density (PSD) they can even account changing detector characteristics. However, such networks requires knowing in...
Reconstructing the structure of thin films and multilayers from measurements scattered x-rays or neutrons is key to progress in physics, chemistry, biology. However, finding all structures compatible with reflectometry data computationally prohibitive for standard algorithms, which typically results unreliable analysis only a single potential solution identified. We address this lack reliability probabilistic deep learning method that identifies realistic seconds, redefining standards...
We demonstrate unprecedented accuracy for rapid gravitational wave parameter estimation with deep learning. Using neural networks as surrogates Bayesian posterior distributions, we analyze eight events from the first LIGO-Virgo Gravitational-Wave Transient Catalog and find very close quantitative agreement standard inference codes, but times reduced O(day) to 20 s per event. Our are trained using simulated data, including an estimate of detector noise characteristics near This encodes signal...
Binary black holes (BBHs) in eccentric orbits produce distinct modulations the emitted gravitational waves (GWs). The measurement of orbital eccentricity can provide robust evidence for dynamical binary formation channels. We analyze 57 GW events from first, second and third observing runs LIGO-Virgo-KAGRA (LVK) Collaboration using a multipolar aligned-spin inspiral-merger-ringdown waveform model with two parameters: relativistic anomaly. This is made computationally feasible...
Deep neural network (DNN) based salient object detection in images on high-quality labels is expensive. Alternative unsupervised approaches rely careful selection of multiple handcrafted saliency methods to generate noisy pseudo-ground-truth labels. In this work, we propose a two-stage mechanism for robust prediction, where the first stage involves refinement pseudo generated from different methods. Each method substituted by deep that learns These are refined incrementally iterations via...
Neural posterior estimation methods based on discrete normalizing flows have become established tools for simulation-based inference (SBI), but scaling them to high-dimensional problems can be challenging. Building recent advances in generative modeling, we here present flow matching (FMPE), a technique SBI using continuous flows. Like diffusion models, and contrast flows, allows unconstrained architectures, providing enhanced flexibility complex data modalities. Flow matching, therefore,...
Abstract We provide a dispersion-theoretical representation of the reaction amplitudes $$\gamma K\rightarrow K \pi $$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>γ</mml:mi> <mml:mi>K</mml:mi> <mml:mo>→</mml:mo> <mml:mi>π</mml:mi> </mml:mrow> </mml:math> in all charge channels, based on modern pion–kaon P -wave phase shift input. Crossed-channel singularities are fixed from phenomenology as far possible. demonstrate how subtraction constants can be matched to...
We study the quark-mass dependence of $$\omega \rightarrow 3\pi $$ decays, based on a dispersion-theoretical framework. rely quark-mass-dependent scattering phase shift for pion–pion P-wave extracted from unitarized chiral perturbation theory. The dispersive representation then takes into account final-state rescattering among all three pions. described formalism may be used as an extrapolation tool lattice QCD calculations three-pion which can serve paradigm case.
Self-attention networks have shown remarkable progress in computer vision tasks such as image classification. The main benefit of the self-attention mechanism is ability to capture long-range feature interactions attention-maps. However, computation attention-maps requires a learnable key, query, and positional encoding, whose usage often not intuitive computationally expensive. To mitigate this problem, we propose novel module with explicitly modeled using only single parameter for low...
Simulation-based inference with conditional neural density estimators is a powerful approach to solving inverse problems in science. However, these methods typically treat the underlying forward model as black box, no way exploit geometric properties such equivariances. Equivariances are common scientific models, however integrating them directly into expressive networks (such normalizing flows) not straightforward. We here describe an alternative method incorporate equivariances under joint...
Organizational buyers are increasingly employing competitive tenders with objective buying criteria to mitigate the influence of personal relationships suppliers and reduce overall cost buying. This paper investigates role salespeople's (i.e., purchasing managers) how they affect supplier selection in such contexts. Drawing on data from 428 across different organizations, this study shows that quality salesperson's relationship buyer influences buyer's evaluation tender proposal, which,...
Atmospheric retrievals (AR) characterize exoplanets by estimating atmospheric parameters from observed light spectra, typically framing the task as a Bayesian inference problem. However, traditional approaches such nested sampling are computationally expensive, thus sparking an interest in solutions based on machine learning (ML). In this ongoing work, we first explore flow matching posterior estimation (FMPE) new ML-based method for AR and find that, our case, it is more accurate than...
We provide a dispersion-theoretical representation of the reaction amplitudes $\gamma K\rightarrow K \pi$ in all charge channels, based on modern pion--kaon $P$-wave phase shift input. Crossed-channel singularities are fixed from phenomenology as far possible. demonstrate how subtraction constants can be matched to low-energy theorem and radiative couplings $K^*(892)$ resonances, thereby providing model-independent framework for future analyses high-precision kaon Primakoff data.
Mergers of binary neutron stars (BNSs) emit signals in both the gravitational-wave (GW) and electromagnetic (EM) spectra. Famously, 2017 multi-messenger observation GW170817 led to scientific discoveries across cosmology, nuclear physics, gravity. Central these results were sky localization distance obtained from GW data, which, case GW170817, helped identify associated EM transient, AT 2017gfo, 11 hours after signal. Fast analysis data is critical for directing time-sensitive observations;...
Reconstructing the structure of thin films and multilayers from measurements scattered X-rays or neutrons is key to progress in physics, chemistry, biology. However, finding all structures compatible with reflectometry data computationally prohibitive for standard algorithms, which typically results unreliable analysis only a single potential solution identified. We address this lack reliability probabilistic deep learning method that identifies realistic seconds, setting new standards...
Inferring atmospheric properties of exoplanets from observed spectra is key to understanding their formation, evolution, and habitability. Since traditional Bayesian approaches retrieval (e.g., nested sampling) are computationally expensive, a growing number machine learning (ML) methods such as neural posterior estimation (NPE) have been proposed. We seek make ML-based (1) more reliable accurate with verified results, (2) flexible respect the underlying networks choice assumed noise models....
Context . Inferring atmospheric properties of exoplanets from observed spectra is key to understanding their formation, evolution, and habitability. Since traditional Bayesian approaches retrieval (e.g., nested sampling) are computationally expensive, a growing number machine learning (ML) methods such as neural posterior estimation (NPE) have been proposed. Aims We seek make ML-based (1) more reliable accurate with verified results, (2) flexible respect the underlying networks choice...
Modern large-scale physics experiments create datasets with sizes and streaming rates that can exceed those from industry leaders such as Google Cloud Netflix. Fully processing these requires both sufficient compute power efficient workflows. Recent advances in Machine Learning (ML) Artificial Intelligence (AI) either improve or replace existing domain-specific algorithms to increase workflow efficiency. Not only the performance of current algorithms, but they often be executed more quickly,...
Ausschreibungsunterlagen werden mangels zeitgerechter Anfechtung auch dann bestandsfest, wenn diese Vergaberechtswidrigkeiten enthalten sollten. Ausdrücklich bestandsfeste Festlegungen in gehen insoweit den vergaberechtlichen Regelungen vor. Das Auslegungskriterium der gesetzeskonformen Interpretation kann nicht dazu verwendet werden, eine nachträgliche Rechtmäßigkeitsprüfung bestandsfester vorzunehmen. Der Verwaltungsgerichtshof hat bei Durchbrechung von Bestandsfestigkeit (sogenannte...
Self-attention networks have shown remarkable progress in computer vision tasks such as image classification. The main benefit of the self-attention mechanism is ability to capture long-range feature interactions attention-maps. However, computation attention-maps requires a learnable key, query, and positional encoding, whose usage often not intuitive computationally expensive. To mitigate this problem, we propose novel module with explicitly modeled using only single parameter for low...