PROJECT TITLE :

A Survey on Large-scale Machine Learning

ABSTRACT:

Text mining, visual classification, and recommender systems are just some of the real-world applications that have made extensive use of machine learning in recent years. Machine learning can provide profound insights into data, enabling machines to make accurate predictions. When working with large amounts of data, however, the majority of sophisticated machine learning approaches incur enormously high time costs. This problem emphasizes the need for large-scale machine learning (LML), which seeks to learn patterns from large amounts of data in an efficient manner while maintaining comparable performance. In this paper, we provide a comprehensive survey on the various LML methods that are currently in use in order to serve as a guide for the upcoming developments in this field. First, we categorize these LML methods according to the various ways in which scalability can be improved. These categorizations are as follows: 1) model simplification with regard to computational complexities; 2) optimization approximation with regard to computational efficiency; and 3) computation parallelism with regard to computational capabilities. After that, we classify the methods in each perspective according to the scenarios that they are designed to address, and then we present representative methods that are in accordance with intrinsic strategies. In the end, we discuss potential directions and open issues that have the potential to be addressed in the future after analyzing their limitations and pointing out their shortcomings.