arXiv subjects

M. Di Giovanni

Publications and source records attributed to M. Di Giovanni.

3 recordsLinked to original sources

arXiv2024

Two sides of the same coin: the F-statistic and the 5-vector method

This work explores the relationship between two data-analysis methods used in the search for continuous gravitational waves in LIGO-Virgo-KAGRA data: the $\mathcal{F}$-statistic and the 5-vector method. We show that the 5-vector method can be derived from a maximum likelihood framework similar to the $\mathcal{F}$-statistic. Our analysis demonstrates that the two methods are statistically equivalent, providing the same detection probability for a given false alarm rate. We extend this comparison to multiple detectors, highlighting differences from the standard approach that simply combines 5-vectors from each detector. In our maximum likelihood approach, each 5-vector is weighted by the observation time and sensitivity of its respective detector, resulting in efficient estimators and analytical distributions for the detection statistic. Additionally, we present the analytical computation of sensitivity for different searches, expressed in terms of the minimum detectable amplitude.

gr-qc↗

arXiv2020

Physical Symmetries Embedded in Neural Networks

Neural networks are a central technique in machine learning. Recent years have seen a wave of interest in applying neural networks to physical systems for which the governing dynamics are known and expressed through differential equations. Two fundamental challenges facing the development of neural networks in physics applications is their lack of interpretability and their physics-agnostic design. The focus of the present work is to embed physical constraints into the structure of the neural network to address the second fundamental challenge. By constraining tunable parameters (such as weights and biases) and adding special layers to the network, the desired constraints are guaranteed to be satisfied without the need for explicit regularization terms. This is demonstrated on upervised and unsupervised networks for two basic symmetries: even/odd symmetry of a function and energy conservation. In the supervised case, the network with embedded constraints is shown to perform well on regression problems while simultaneously obeying the desired constraints whereas a traditional network fits the data but violates the underlying constraints. Finally, a new unsupervised neural network is proposed that guarantees energy conservation through an embedded symplectic structure. The symplectic neural network is used to solve a system of energy-conserving differential equations and out-performs an unsupervised, non-symplectic neural network.

physics.comp-ph↗

arXiv2018

Calibration of Advanced Virgo and Reconstruction of the Gravitational Wave Signal h(t) during the Observing Run O2

In August 2017, Advanced Virgo joined Advanced LIGO for the end of the O2 run, leading to the first gravitational waves detections with the three-detector network. This paper describes the Advanced Virgo calibration and the gravitational wave strain h(t) reconstruction during O2. The methods are the same as the ones developed for the initial Virgo detector and have already been described in previous publications, this paper summarizes the differences and emphasis is put on estimating systematic uncertainties. Three versions of the h(t) signal have been computed for the Virgo O2 run, an online version and two post-run reprocessed versions with improved detector calibration and reconstruction algorithm. A photon calibrator has been used to establish the sign of h(t) and to make an independent partial cross-check of the systematic uncertainties. The uncertainties reached for the latest h(t) version are 5.1% in amplitude, 40 mrad in phase and 20 microseconds in timing.

gr-qc↗