Online Program

Return to main conference page
Thursday, May 17
Computing Science
Automated Model Building
Thu, May 17, 10:30 AM - 12:00 PM
Grand Ballroom D
 

D3M Automated Model Building and Diagnostics (304683)

*Curtis Lisle, KnowledgeVis LLC 

Keywords: Automated Data Analysis, Model Selection, Model Diagnostics, Subject Matter Expert, Domain-Knowledge Data-Science Integration

DARPA's Data Driven Discovery of Models (D3M) program is an aggressive effort with the goal to design automated data science tools for use directly by domain experts without substantial statistics or data science experience themselves. We will cover the goals, approach, and early results of this program. The architecture for a general-purpose data analysis tool has been subdivided into three separate areas: The first area is a library of best-in-class algorithms that can be composed into solution pipelines. The second area focuses on systems that attempt automated model comparison and model selection, given an input dataset and a problem or hypothesis to explore. These systems will develop suggested models. The third and final area is the interactive user interfaces that present the model details and results for user review and action. We will explore each area of this architecture and discuss how models can be developed and evaluated by the non-data scientist, domain experts using this approach. A live demonstration of a prototype D3M system will be included in this presentation.