Window function convolution with deep neural network models

Traditional estimators of the galaxy power spectrum and bispectrum are sensitive to survey geometry. They yield spectra that differ from true underlying signal since they convolved with window function survey. For current future generations experiments, this bias is statistically significant on large scales. It thus imperative effect summary statistics distribution accurately modelled. Moreover, operation must be computationally efficient in order allow sampling posterior probabilities while performing Bayesian estimation cosmological parameters. In satisfy these requirements, we built a deep neural network model emulates convolution function, show it provides fast accurate predictions. We trained (tested) using suite 2000 (200) models within cold dark matter scenario, demonstrate its performance agnostic precise values all cases, for better than 0.1% timescale 10 ?s.