arXiv subjects

Robin Geyer

Publications and source records attributed to Robin Geyer.

3 recordsLinked to original sources

arXiv2026

Gravitational waves and small-field astrometry

Astrometric observations can, in principle, be used to detect gravitational waves. In this paper we give a practical overview of the gravitational wave effects which can be expected specifically in small-field astrometric data. Particular emphasis is placed on the differential effect between pairs of sources within a finite field of view. We also present several general findings that are not restricted to the small-field case. A detailed theoretical derivation of the general astrometric effect of a plane gravitational wave is provided. Numerical simulations, which underline our theoretical findings, are presented. We find that small-field missions suffer from significant detrimental properties, largely because their relatively small fields only allow the measurement of small differential effects which can be expected to be almost totally absorbed by standard plate calibrations.

gr-qc↗

arXiv2023

Invariant Anomaly Detection under Distribution Shifts: A Causal Perspective

Anomaly detection (AD) is the machine learning task of identifying highly discrepant abnormal samples by solely relying on the consistency of the normal training samples. Under the constraints of a distribution shift, the assumption that training samples and test samples are drawn from the same distribution breaks down. In this work, by leveraging tools from causal inference we attempt to increase the resilience of anomaly detection models to different kinds of distribution shifts. We begin by elucidating a simple yet necessary statistical property that ensures invariant representations, which is critical for robust AD under both domain and covariate shifts. From this property, we derive a regularization term which, when minimized, leads to partial distribution invariance across environments. Through extensive experimental evaluation on both synthetic and real-world tasks, covering a range of six different AD methods, we demonstrated significant improvements in out-of-distribution performance. Under both covariate and domain shift, models regularized with our proposed term showed marked increased robustness. Code is available at: https://github.com/JoaoCarv/invariant-anomaly-detection.

cs.LG↗

arXiv2017

Investigation of Algorithms for Highly Nonlinear Model Fitting on Big Datasets

This thesis investigates algorithms regarding their applicability for highly nonlinear model fitting on big datasets. Various mathematical methods are presented with which a model fit using the least squares criterion is possible. Special requirements regarding the processing of large data sets as a basis for such a model fit are discussed. The specific example of the search for gravitational wave signals in simulated data of the ESA satellite mission Gaia is used to demonstrate how a model fit is possible, even with complex models and large amount of data. For this purpose, a highly parallel prototype of a future search software is implemented. The resulting prototype uses a hybrid algorithm which utilizes a linear search, an evolutionary algorithm and a classical iterative Gauss-Newton fit. The performance and behavior of its components are investigated in detail. With the help of software presented in this work it has been possible for the first time to detect gravitational wave signals in simulated astrometric data, and to determine their parameters. Furthermore, it can be concluded from the runtime behavior of the software that such a search is also possible in real data of the Gaia mission.

astro-ph.IM↗