Source author record

Manuel D. Morales

Manuel D. Morales appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

gr-qc astro-ph.IM astro-ph.SR Machine Learning physics.data-an

Catalog footprint

What is connected

3works

5topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2021arXiv

Using supervised learning algorithms as a follow-up method in the search of gravitational waves from core-collapse supernovae

We present a follow-up method based on supervised machine learning (ML) to improve the performance in the search of gravitational wave (GW) burts from core-collapse supernovae (CCSNe) using the coherent WaveBurst (cWB) pipeline. The ML model discriminates noise from signal events using as features a set of reconstruction parameters provided by cWB. Detected noise events are discarded yielding to a reduction of the false alarm rate (FAR) and of the false alarm probability (FAP) thus enhancing of the statistical significance. We tested the proposed method using strain data from the first half of the third observing run of advanced LIGO, and CCSNe GW signals extracted from 3D simulations. The ML model is learned using a dataset of noise and signal events, and then it is used to identify and discard noise events in cWB analyses. Noise and signal reduction levels were examined in single detector networks (L1 and H1) and two detector network (L1H1). The FAR was reduced by a factor of $\sim10$ to $\sim100$, there was an enhancement in the statistical significance of $\sim1$ to $\sim2σ$, while there was no impact in detection efficiencies.

preprint2020arXiv

Deep learning for gravitational-wave data analysis: A resampling white-box approach

In this work, we apply Convolutional Neural Networks (CNNs) to detect gravitational wave (GW) signals of compact binary coalescences, using single-interferometer data from LIGO detectors. As novel contribution, we adopted a resampling white-box approach to advance towards a statistical understanding of uncertainties intrinsic to CNNs in GW data analysis. Resampling is performed by repeated $k$-fold cross-validation experiments, and for a white-box approach, behavior of CNNs is mathematically described in detail. Through a Morlet wavelet transform, strain time series are converted to time-frequency images, which in turn are reduced before generating input datasets. Moreover, to reproduce more realistic experimental conditions, we worked only with data of non-Gaussian noise and hardware injections, removing freedom to set signal-to-noise ratio (SNR) values in GW templates by hand. After hyperparameter adjustments, we found that resampling smooths stochasticity of mini-batch stochastic gradient descend by reducing mean accuracy perturbations in a factor of $3.6$. CNNs were quite precise to detect noise but not sensitive enough to recall GW signals, meaning that CNNs are better for noise reduction than generation of GW triggers. However, applying a post-analysis, we found that for GW signals of SNR $\geq 21.80$ with H1 data and SNR $\geq 26.80$ with L1 data, CNNs could remain as tentative alternatives for detecting GW signals. Besides, with receiving operating characteristic curves we found that CNNs show much better performances than those of Naive Bayes and Support Vector Machines models and, with a significance level of $5\%$, we estimated that predictions of CNNs are significant different from those of a random classifier. Finally, we elucidated that performance of CNNs is highly class dependent because of the distribution of probabilistic scores outputted by the softmax layer.

preprint2015arXiv

Quasi-normal acoustic oscillations in the Michel flow

We study spherical and nonspherical linear acoustic perturbations of the Michel flow, which describes the steady radial accretion of a perfect fluid into a nonrotating black hole. The dynamics of such perturbations are governed by a scalar wave equation on an effective curved background geometry determined by the acoustic metric, which is constructed from the spacetime metric and the particle density and four-velocity of the fluid. For the problem under consideration in this article the acoustic metric has the same qualitative features as an asymptotically flat, static and spherically symmetric black hole, and thus it represents a natural astrophysical analogue black hole. As for the case of a scalar field propagating on a Schwarzschild background, we show that acoustic perturbations of the Michel flow exhibit quasi-normal oscillations. Based on a new numerical method for determining the solutions of the radial mode equation, we compute the associated frequencies and analyze their dependency on the radii of the event and sonic horizons and the angular momentum number. Our results for the fundamental frequencies are compared to those obtained from an independent numerical Cauchy evolution, finding good agreement between the two approaches. When the radius of the sonic horizon is large compared to the event horizon radius, we find that the quasi-normal frequencies scale approximately like the surface gravity associated with the sonic horizon.