Data-Driven Discovery of Models (D3M)

Abstract

The Data-Driven Discovery of Models (D3M) program, expanding on technical opportunities emerging from the XDATA program, is developing automated model discovery techniques and tools that enable non-expert users to create empirical models of real, complex processes and phenomena. The ability to understand the battlespace is driven increasingly by analysis of sensor and open source data, and the construction of empirical models that enable decision makers to predict behaviors and anticipate contingencies during tactical and strategic planning. The DoD and the Intelligence Community (IC) are fundamentally limited in this regard by a shortage of expert data scientists. D3M will address this need by creating technologies that automate the construction of complex empirical models. D3M technologies will include a library of data modeling primitives that is automatically selectable, given data and an outcome; automated approaches for composition of complex models from modeling primitives; and intuitive mechanisms for human-model interaction that enable curation of models by non-experts. D3M technical development will focus on the types of empirical modeling problems commonly encountered by the DoD and IC.

Document Details

Document Type
Accomplishment
Publication Date
Oct 01, 2018
Source ID
0cb74bf3982ad6946fafb7cace50f75f

Tags

Fields of Study

  • Computer science
  • Engineering

Readers

  • Computational Modeling and Simulation
  • Database Systems and Applications
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Related Documents