Sparse, deep neural networks for the early detection of gravitational waves from binary neutron stars
Summary
In the field of gravitational-wave detection, machine learning has become a part of the landscape. In this work, we build upon a previous work, which gave a proof of concept of the use of convolutional neural network in the detection of the early phases of an inspiraling binary neutron star. Obtaining such early detection would allow one to send out alerts to astronomers, so that other phases of the merger can be observed using other messengers.
In this work, we adapt two of the latest techniques from other neural network fields, called inception modules, adapted from Google Inception-Resnet-v2 and dilation, as implemented in Wavenet. Here, for the first time, we revise those tools to be used for one dimensional data. We show they can lead to improvements in the early detection of binary neutron stars.