Central versus Distributed Statistical Computing Algorithms-A Comparison

Abstract

Distributed statistical learning algorithms are performing many machine learning tasks in a distributed environment. Some scenarios where data sharing is desired among many parties and it may need to increase the efficiency and statistical accuracy of the underlying algorithms. Due to the increase in the size and complexity of today’s big data, it is very important to solve problems with a very large number of features, records, and training samples. As a result, it is necessary to deal with the distributed transfer of these datasets as well as their underlying distributed solution methods efficiently and effectively. This paper compares the efficiency and accuracy of a distributed statistical method with a central method with simple regression and classification algorithms.