I-Corps: Factor graph computing for data-driven decision-making

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is to democratize the power of machine learning and predictive analytics to help non-engineering business personnel realize the full value of their data. Enabling this will drive better organizational decision making at all levels and the creation of additional business value for commercial and governmental organizations of all kinds, and of all sizes. Data-driven decisions using big data analytics holds massive promise, but has largely remained unfulfilled primarily because, in its current form, it is unaffordable to all but the most technologically advanced organizations. Data-driven businesses have 5-6% higher productivity and can potentially add $3 Trillion in value globally, per year. A wide range of sectors, such as manufacturing, retail, finance, healthcare, security, and governmental services will see significant commercial impact including higher productivity, better utilization of resources, and the acceleration of the deployment of new products, services, and technologies. By enabling commercial and governmental entities to more effectively utilize their data and resources, businesses and government agencies will increase their productivity, more effectively utilize their personnel, improve their competitiveness, and eliminate waste resources while increasing the flow of products and services that benefit society at large. This I-Corps project is based on ground breaking technology meant to realize a scalable, flexible and easy-to-use data-processing infrastructure. At the highest level, building such a platform requires: (a) coming up with the right abstraction or language that accommodates all sorts of computation at scale; (b) implementing the architecture to realize such computation at scale; (c) the ability to go from "sandboxing" or "prototyping" to a production environment instantly; and (d) the ability to work in heterogenous data environments instantly. We have put forth a novel computational language, called factor graph computing. Such a computation framework is "Turing complete". Factor graph computing allows for performing data transformation, predictive modeling, and optimization at scale to enable data-driven decisions. Such a platform eliminates the need for data engineering, provides flexibility to build predictive models instantly, allows for seamless evolution by bringing in new datasets in the mix, and requires minimal support to maintain the infrastructure and allow for generating prescriptive decisions at scale. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.