Advancing Atomic Modeling: Integration of Computational Clusters & Neural Network Techniques

High-precision atomic structure calculations require accurate modelling of electronic correlations involving large multielectron atomic structures. Here, we develop a deep-learning methodology that enables the preselection of the most pertinent configurations from huge basis sets until the desired precision is attained based on a weighted scale. Our approach performs under the control of a convolutional neural network that is being trained by previous GRASP examples. The findings for a number of cases involving numerous electron atoms demonstrate that deep learning can greatly reduce the amount of computer memory and processing time needed and makes large-scale calculations on previously inaccessible basis sets conceivable. Through the making of a cluster made of recycled Dell CPUs and the installation of the Linux operating system Rocky Linux 8.8 the cluster has the ability of preforming numerous GRASP examples which allow us to have confidence that the neural net is going to preform and were able to achieve our goal.