Identifying extra high frequency gravitational waves generated from oscillons with cuspy potentials using deep neural networks

Abstract

During oscillations of cosmology inflation around the minimum of a cuspy potential after inflation, the existence of extra high frequency gravitational waves (HFGWs) (∼GHz) has been proven effectively recently. Based on the electromagnetic resonance system for detecting such extra HFGWs, we adopt a new data processing scheme to identify the corresponding GW signal, which is the transverse perturbative photon fluxes (PPF). In order to overcome the problems of low efficiency and high interference in traditional data processing methods, we adopt deep learning to extract PPF and make some source parameters estimation. Deep learning is able to provide an effective method to realize classification and prediction tasks. Meanwhile, we also adopt anti-overfitting technique and make adjustment of some hyperparameters in the course of study, which improve the performance of classifier and predictor to a certain extent. Here the convolutional neural network (CNN) is used to implement deep learning process concretely. In this case, we investigate the classification accuracy varying with the ratio between the number of positive and negative samples. When such ratio exceeds to 0.11, the accuracy could reach up to 100%. Besides, we also investigate the classification accuracy with different amplitude of extra HFGWs. As a predictor, the mean relative error of parameters estimation decreases when the amplitude of extra HFGWs increases. Especially, when amplitude h(t) is in 10 −31 -10 −30 the mean relative error reaches around 0.014. On the contrary, the mean relative error increases with frequency increasing in 10 8 -10 11 Hz. At the optimal resonance frequency 5×10 9 Hz, the mean relative error is 0.12. Then we also study the mean relative error varying with waist radius W 0 of Gaussian beam, its optimal value is 0.138 when W 0 is in (0.05 m, 0.1 m) approximately. Compared with classifiers and predictors using other machine learning algorithms, deep CNN for our datasets has higher accuracy and lower error.

Data classification and parameter estimations with deep learning to the simulated time-domain high-frequency gravitational waves detections

“…Also, λ e , κ e , ω e = κ e c = 2π f e , z 0 = π W 2 0 /λ e and δ are the wavelength, wave vector, center frequency, the Rayleigh length, and the phase factor of the Gaussian beam, respectively. In this work, we specifically take that [33,57], ψ 0 = 1260 (the corresponding power of the used laser is about 10 W), δ = 1.9π, and W 0 = (0.05, 0.1)m. The high stationary magnetic field is assumed to be distributed in the region: [45]. Formally, a modulated function exp −(t − β/f e ) 2 /f α e is introduced to extend the detection bandwidth, wherein the modulated parameter is taken as: α = −1.95 and β = 5.0.…”

“…Certainly, the detection of the HFGWs is not easy, as their response signals are significantly weak, while the background noises are significantly strong [33]. Therefore, besides the installations are expected to be built for the experimental detections, various effective big data processing approaches are also required to be developed for the abundant raw data (with strong noise) processings, typically including the matched filtering algorithms [34] used in LIGO-Virgo observations [2].…”

“…Given the deep neural network algorithm has been successfully utilized to implement the identification of the simulatively detected data of the HFGWs in the spatial domain [33], here we further develop an effective deep learning method to identify the simulatively detected data in the time domain. The simulatively detected signal is generated by the first-order transverse EM responses in the Li-Baker detector [45,46] but can be easily generated in the other potential HFGW detectors.…”

Gravitational wave signal recognition of O1 data by deep learning

“…In the past few years, some active researchers have demonstrated empirical successes of deep neural networks in the applications to data analysis of gravitational waves [41][42][43][44][45]. These active researchers include George and his coworkers [9,10], Gabbard and his coworkers [46], and others [11,12,47]. These published works used the convolutional neural network (CNN) from different perspectives to identify GW signals with low signal-to-noise ratio (SNR).…”

Attention-mechanism–based network characteristic analysis for major depressive disorder detection

“…By setting an appropriate convolution kernel, CNN can effectively learn grid-like data. Many studies have proved that this end-to-end learning scheme can effectively extract complex nonlinear relationship [39][40][41][42][43]. So far, CNN-based researches have achieved many good results in EEG signal analysis [44][45][46].…”