Evolution of the cosmic microwave background (CMB) radiation temperature is investigated in the context of modified generalized Chaplygin gas (mgCG) model with a Deep Neural Network (DNN) analysis. First, we reconstruct a new formulation for the temperature-red shift relation of the CMB photons and then find out the best-fitting values of the arbitrary parameters introduced in the model via the help of a neural network mechanism including the genetic algorithm (GA) and the Fisher information matrix (FIM) approach (we name this mechanism as the genetic Fisher algorithm or shortly the GFA) by considering astrophysical measurements given in literature for some galaxy cluster and quasar samples. Subsequently, we apply the long short-term memory plus Dropout (LSTMD) neural network approach, which is emerging as a solution for problematic issues in the DNN investigations (e.g. the over-fitting), to well-known astrophysical parameters. In view of the presented conclusions, the whole mechanism comes forward as a new up and coming step on how one can cultivate a new neural network that can calculate its own confidence region. In addition to this, it is significant to emphasize here that the promising approach helps to decrease the computational load of pricey codes for astrophysical models and probes the necessity of alternative CMB radiation temperature models. As a result, we prove that the fitted form of our model is in very good agreement with the observational data. • We constructed a new theoretical model for the evolution CMB radiation temperature. • New model was fitted by making use of the recent observational data. • The fitting process was improved with the help of deep learning analysis. • The fitted model agrees with the current experimental data and the Λ CDM cosmology. • We showed that deep learning algorithms can take crucial roles in astrophysics.
Read full abstract