The Defense Advanced Research Projects Agency will host a proposer’s day in search of expertise to support Learning with Less Label, a program aiming to reduce amounts of information needed to train machine learning models.
The event will run on July 12 at the DARPA Conference Center in Arlington, Va., the agency said Wednesday.
“Under LwLL, we are seeking to reduce the amount of data required to build a model from scratch by a million-fold, and reduce the amount of data needed to adapt a model from millions to hundreds of labeled examples,” said Wade Shen, a program manager in DARPA’s Information Innovation Office, and head of the LwLL program.
The program will focus on two study areas, the first of which is building algorithms that learn and adapt with a reduced amount of labeled examples.
The second area is identifying decision difficulty and data complexity problems found in machine learning.