An Advanced Methodological Optimization of Stochastic Gradient Descent for High-Dimensional Big Data Analysis in Computational Systems

Abstract

The acceleration of high-dimensional big data analysis is paramount in the evolution of computational science. This paper presents an advanced optimization technique for the stochastic gradient descent (SGD) algorithm, fine-tuned for efficiency in complex datasets. Leveraging adaptive learning rates and momentum tuning, our method significantly reduces convergence time while maintaining accuracy. Experimental results demonstrate enhanced performance across various computational platforms, establishing a new benchmark for data-driven decision-making processes.