Speaker
Description
Line-of-sight velocity disturbs the observed spatial positions of galaxies, thereby introducing anisotropies in their observed distribution, which is known as redshift space distortions (RSD). This complication inevitably affects the accurate extraction of important feature of the density field, such as baryonic acoustic oscillations (BAO). In this work, we propose a new scheme based on artificial neural networks (ANN), to directly estimate the line-of-sight velocities of individual galaxies from an observed redshift space galaxy distribution. By training the network with environmental characteristics surrounding each galaxy in redshift space, our ANN model can recover the line-of-sight velocity of each individual galaxy accurately. When using this velocity to eliminate the RSD effect, the two-point correlation function (TPCF) in real space can be recovered with better than 1% accuracy at $s$ > 8 $h^{-1}\mathrm{Mpc}$, with only 4% deviation over all scales compared to the ground truth. The real-space power spectrum can be recovered with better than 3% deviation on $k$< 0.5 $\mathrm{Mpc}^{-1}h$, and less than 5% for all $k$-modes. The quadrupole moment of the TPCF or power spectrum is almost zero down to $s$ = 10 $h^{-1}\mathrm{Mpc}$ or all $k$-modes, in excellent agreement with the expected results in the real space, which demonstrates an efficient correction of the spatial anisotropy introduced by the RSD effect. Although our ANN is trained using simulation data with one cosmological model, tests with other cosmological models indicate a weak cosmology dependency. Our scheme offers a novel avenue to predict the peculiar velocity of galaxies, to eliminate the RSD effect directly in future large galaxy surveys, and to reconstruct the 3-D cosmic velocity field accurately.
