The Practical Approaches of Datasets in Machine Learning

The machine learning research is primarily organized around benchmark datasets. They serve as a means of measuring progress toward shared objectives and unite scholars around common research concerns. Despite the crucial function of dynamics of benchmark dataset use and reuse in the field of machine learning allows the field of using benchmark datasets in reality. This paper delves into these dynamics of the usage of datasets in machine learning. Within task communities, this work has been observed as a growing concentration on a small number of datasets, a large adoption of datasets from other tasks, and a concentration across the field on datasets provided by academics based in a few groups of prestigious institutions.