A novel radial basis Bayesian regularization deep neural network for the Maxwell nanofluid applied on the Buongiorno model

The aim of this work is to provide the numerical solutions of the fluid model by using the stochastic computing paradigms. The linear/exponential stretching sheets on magneto-rotating flow based on the Maxwell nanofluid have been provided using the Buongiorno model with the impacts of uneven heat source/sink, varying thermal conductivity and reactive species. The solutions of this transformed ordinary differential exponential stretching sheet model have been presented using a novel ‘radial basis’ (RB) activation function together with the Bayesian regularization deep neural network (BRDNN), i.e., RB-BRDNN. The deep neural network is presented into two hidden layers, while thirteen and twenty-five numbers of neurons have been used in the first and second layer. A reference dataset is proposed using the Runge-Kutta scheme for the model. The correctness of the stochastic RB-BRDNN procedure is examined through the comparison of proposed and database results, whereas minimal absolute error values provide the accuracy of the scheme. The reliability and competence of the computing RB-BRDNN solver is authenticated using the state transitions, correlation, regression, and error histograms.