Title: Reconstructing the Hubble diagram of gamma-ray bursts using deep learning

Abstract: We calibrate the distance and reconstruct the Hubble diagram of gamma-ray bursts (GRBs) using deep learning. We construct an artificial neural network, which combines the recurrent neural network and Bayesian neural network, and train the network using the Pantheon compilation of type-Ia supernovae. The trained network is used to calibrate the distance of 174 GRBs based on the Combo-relation. We verify that there is no evident redshift evolution of Combo-relation, and obtain the slope and intercept parameters, $\gamma=0.856^{+0.083}_{-0.078}$ and $\log A=49.661^{+0.199}_{-0.217}$, with an intrinsic scatter $\sigma_{\rm int}=0.228^{+0.041}_{-0.040}$. Our calibrating method is independent of cosmological model, thus the calibrated GRBs can be directly used to constrain cosmological parameters. It is shown that GRBs alone can tightly constrain the $\Lambda$CDM model, with $\Omega_{\rm M}=0.280^{+0.049}_{-0.057}$. However, the constraint on the $\omega$CDM model is relatively looser, with $\Omega_{\rm M}=0.345^{+0.059}_{-0.060}$ and $\omega<-1.414$. The combination of GRBs and Pantheon can tightly constrain the $\omega$CDM model, with $\Omega_{\rm M}=0.336^{+0.055}_{-0.050}$ and $\omega=-1.141^{+0.156}_{-0.135}$.