The U.S. Defense Department's research and development arm is offering to fund projects that will simplify the massively complex task of building models for machine learning applications.
Models are a fundamental part of machine learning. Similar to algorithms, they help teach computers to, say, identify a cat in a photo, forecast weather from historical data or sort spam from legitimate email.
But writing the models takes time and requires many skills. Typically, data scientists, subject matter experts and software engineers all have to come together to develop the model.
When New York University researchers wanted to model block-by-block traffic flow data for the city, it took 60 person-months of work by data scientists to prepare the data for use and an additional 30 person-months to develop the model.
The Defense Advanced Research Projects Agency (DARPA) wants to change that.
It's proposing research into "automated model discovery systems" that would allow a subject matter expert with no data science expertise to create a model.
The Department of Defense sees great potential for machine learning, especially as the volume of open-source data available continues to increase year after year.
DARPA said it envisages models that help decision makers predict behavior, such as the movement of enemy troops during a conflict, and envisage scenarios, such as weather and traffic. It could even be used to figure out where to best place personnel within the DOD.
Driving the project is not just the need to save time and money. DARPA cites predications that the world could be short of up to 180,000 data scientists this year alone, and that shortfall could get worse in coming years.
The Data Driven Discovery of Models (D3M) project will take place over two phases of 24 months each. More details can be found on the Fed Biz Opps website.