Optimizing the Number of Nodes in Artificial Intelligence Neural Network Using Parallel Computing Methods

Abstract

Determining the optimal configuration/ architecture for neural network could be a very time-consuming process. It includes going through big number of scenarios to simulate and validate different architecture. This paper will go over a process for intelligently automating this tedious task by providing a utility that will automatically validate the different permutation and picking the optimum configuration. This is done with maximum efficiency by utilizing parallel computing methods combined with in-memory tables to produce very fast result. The new methods enabled Exploration and Producing (E&P) data scientists to simulate and validate a big number of neural network configurations/ architectures in shorter time, allowing them to identify the optimum neural network configurations that can generate the best prediction results. Leading to more accurate Artificial Intelligent/ Machine Learning (AI/ML) models. To quantify this method, this paper approach can cut the time needed for such process from weeks to hours, especially when dealing with modeling E&P real-time problem. The process involved utilizing parallel computing methods combined with in-memory tables, not only in processing complex Artificial Intelligent/ Machine Learning (AI/ML) models, but also in developing the complex models. This approach is intended to empower data scientist in the Exploration and Producing field to quickly reach the most optimize model, leading to faster approach for safer, and more cost-effective solutions