luminosity distance reconstruction using an ensemble of neural networks

Exploring the consistency of a dataset with the λcdm model across low and high redshifts stands as a compelling topic in cosmology. given the capability of neural networks to reconstruct an unknown function, we employed an ensemble of neural networks to reconstruct the luminosity distance based on the pantheon+ dataset. each network in the ensemble consists of various numbers of layers and neurons. since the neural network can easily provide a reconstruction with a small value of χ2, it is possible to find a reconstruction with χ^2 smaller than the standard λcdm. we selectively choose those reconstructions with a χ^2 value smaller than the best-fit λcdm model. our findings reveal that all reconstructions yield a smaller luminosity distance at high redshifts compared to the best λcdm. assuming a flat universe, we transformed the reconstructions into the hubble parameter as a function of redshifts and compared the results with predictions of the λcdm model.