A distributed and unified API service for machine learning models

Abstract

Machine learning models are usually based on a single dataset and a single architecture approach. But this approach falls behind in many practical scenarios. Many techniques such as boosting and bagging have been developed in the past that enables better predictions by ensembling multiple models. In this paper, we propose an extension to the existing ensembling techniques by developing a community-driven system of distributed API that unifies several models to produce an ensembling effect along with several other benefits such as wider availability, greater usability, lesser resource constraints and an infrastructure for future developments in the field.