Optimised neural network predictions of galaxy formation histories using semi-stochastic corrections

We present a novel methodology to improve predictions of galaxy formation histories by incorporating semi-stochastic corrections to account for short-timescale variability. Our paper addresses limitations in existing models that capture broad trends in galaxy evolution, but fail to reproduce the bursty nature of star formation and chemical enrichment, resulting in inaccurate predictions of key observables such as stellar masses, optical spectra, and colour distributions. We introduce a simple technique to add a stochastic components by utilizing the power spectra of galaxy formation histories. We justify our stochastic approach by studying the correlation between the phases of the halo mass assembly and star-formation histories in the IllustrisTNG simulation, and we find that they are correlated only on timescales longer than 6 Gyr, with a strong dependence on galaxy type. We demonstrate our approach by applying our methodology to the predictions on a neural network trained on hydrodynamical simulations, which failed to recover the high-frequency components of star-formation and chemical enrichment histories. Our methodology successfully recovers realistic variability in galaxy properties at short timescales. It significantly improves the accuracy of predicted stellar masses, metallicities, spectra, and colour distributions and provides a practical framework for generating large, realistic mock galaxy catalogs, while also enhancing our understanding of the complex interplay between galaxy evolution and dark matter halo assembly.