Neural network predictions of (α, n) reaction cross sections at 18.5±3 MeV using the Levenberg-Marquardt algorithm

Abstract

In recent developments, artificial neural networks (ANNs) have demonstrated their capability to predict reaction cross-sections based on experimental data. Specifically, for predicting (a,n) reaction cross-sections, we meticulously fine-tuned the neural network’s performance by optimizing its parameters through the Levenberg-Marquardt algorithm. The effectiveness of this approach is corroborated by notable correlation coefficients; an R-value of 0.90928 for overall correlation, 0.98194 for validation, 0.99981 for testing, and 0.94116 for the comprehensive network prediction. We conducted a rigorous comparison between the results and theoretical computations derived from the TALYS 1.95 nuclear code to validate the predictive accuracy. The mean square error value for artificial neural network results is 7620.92, whereas for TALYS 1.95 calculations, it has been found to be 50,312.74. This comprehensive evaluation process validates the reliability of the ANN based on the Levenberg-Marquardt algorithm in approximating the reaction sections, thus demonstrating its potential for comprehensive investigations. These recent developments confirm the feasibility of using ANN models to gain insight into (a,n) reaction cross-sections