Reconstructing Cosmic Observations with Artificial Neural Networks: A Test of Hubble Constant

Artificial neural networks (ANNs) have become increasingly popular tools in recent years. They are capable of identifying underlying relationships within input data and can reconstruct the distribution of observed data without relying on any a priori cosmological assumptions. In this paper, we utilized ANNs to train simulated data and subsequently constrained the Hubble constant under both the standard cosmological model and the wCDM model. We also compared the constraints obtained from the original data with those from the simulated data to discern any differences. Additionally, we trained a high-redshift sample based on 1048 supernovae (SNe) data and 174 gamma-ray bursts (GRBs) data. Under the standard cosmological model, this sample yielded a constraint on the Hubble constant of \({H}_0 = 66.17 \pm 0.41\). We also discovered that when integrating two datasets with significantly different sample sizes, ANNs effectively retain the characteristic information of the sample with a smaller size.