Gravitational waves produced from the merger of binary neutron stars (BNSs) are accompanied by electromagnetic counterparts, making it possible to identify the associated host galaxy. We explore how properties of the host galaxies relate to the astrophysical processes leading to the mergers. It is thought that the BNS merger rate within a galaxy at a given epoch depends primarily on the galaxy's star-formation history as well as the underlying merger time-delay distribution of the binary systems. The stellar history of a galaxy, meanwhile, depends on the cosmological evolution of the galaxy through time, and is tied to the growth of structure in the Universe. We study the hosts of BNS mergers in the context of structure formation by populating the Universe Machine simulations with gravitational-wave events~ according to a simple time-delay model. We find that different time-delay distributions predict different properties of the associated host galaxies, including the distributions of stellar mass, star-formation rate, halo mass, and local and large-scale clustering of hosts. BNSs that merge today with short delay times prefer to be in hosts that have high star-formation rates, while those with long delay times live in dense regions within massive halos that have low star formation. We show that with ${\mathcal O}(10)$ events from current gravitational-wave detector networks, it is possible to make preliminary distinctions between formation channels which trace stellar mass, halo mass, or star-formation rate. We also find that strategies to follow up gravitational-wave events with electromagnetic telescopes can be significantly optimized using the clustering properties of their hosts.

In this study, the fundamental stellar atmospheric parameters (Teff, log g, [Fe/H] and [{\alpha}/Fe]) were derived for low-resolution spectroscopy from LAMOST DR5 with Generative Spectrum Networks (GSN). This follows the same scheme as a normal artificial neural network with stellar parameters as the input and spectra as the output. The GSN model was effective in producing synthetic spectra after training on the PHOENIX theoretical spectra. In combination with Bayes framework, the application for analysis of LAMOST observed spectra exhibited improved efficiency on the distributed computing platform, Spark. In addition, the results were examined and validated by a comparison with reference parameters from high-resolution surveys and asteroseismic results. Our results show good consistency with the results from other survey and catalogs. Our proposed method is reliable with a precision of 80 K for Teff, 0.14 dex for log g, 0.07 dex for [Fe/H] and 0.168 dex for [{\alpha}/Fe], for spectra with a signal-to-noise in g bands (SNRg) higher than 50. The parameters estimated as a part of this work are available at http://paperdata.china-vo.org/GSN_parameters/GSN_parameters.csv.