Deep learning approach to Hubble parameter

Abstract The main purpose of this work is to show that machine learning algorithms (MLAs) can be used to improve the abilities of cosmological models and to make meaningful astrophysical predictions. As a preliminary step, we construct an expression for the Hubble parameter in the caloric variable Chaplygin gas (cVCG) framework including a particle creation scenario. Then, making use of a set of updated observational data, we obtain the best-fitting values of the auxiliary model parameters. In the main part of the article, we discuss the model from the machine learning (ML) perspective via two different supervised learning-training algorithms: Long Short-Term Memory (LSTM) cells with Dropout and the Fisher Information Matrix (FIM). We see that the constructed theoretical ground yields very successful results when it is used in the MLAs for training and the reliability level of deep learning (DL) analysis is above % 93.