One parameter binary black hole inverse problem using a sparse training set

摘要

In this paper, we use Artificial Neural Networks (ANNs) to estimate the mass ratio q in a binary black hole collision out of the gravitational wave (GW) strain. We assume the strain is a time series (TS) that contains a part of the orbital phase and the ring-down of the final black hole. We apply the method to the strain itself in the time domain and also in the frequency domain. We present the accuracy in the prediction of the ANNs trained with various values of signal-to-noise ratio (SNR). The core of our results is that the estimate of the mass ratio is obtained with a small sample of training signals and resulting in predictions with errors of the order of 1% for our best ANN configurations.