Evaluation of differential cross section data of alpha induced reactions using Bayesian neural network

Alpha-induced nuclear reactions are crucial to understand nucleosynthesis effectively. Differential cross-sections provide us insight into the reaction mechanisms happening during the reaction. Such cross-sections are also helpful in effectively determining the respective reaction rates during nucleosynthesis. In this study, we are using neural networks to obtain the evaluated data for alpha-induced reactions. Since neural networks do not provide the uncertainties in their predictions, we have used Bayesian neural networks to estimate the aleatoric uncertainties. We have used ¹²C(α,el)¹²C reaction as the test reaction. We have developed a multi-fidelity Bayesian neural network to benefit from the theoretical model prediction and available experimental data. We used theoretical estimation of the differential cross-section from TALYS and experimental data from the EXFOR data library. The result obtained from the multi-fidelity Bayesian neural network trained for 18000 data points is shown in the article