Charles River Analytics has partnered with academic institutions to help the Defense Advanced Research Projects Agency boost the efficiency of machine-learning model development.
The company said Tuesday it will work with the University of British Columbia, Northeastern University and University of California Irvine to apply probabilistic programming under DARPA’s Learning with Less Labels effort.
The DARPA-led program aims to reduce data requirements for model production and streamline the process of making these models.
Work under the effort supports CRA’s Probabilistic Label-Efficient Deep Generative Structures or PLEDGES approach. PLEDGES aims to develop analytical technologies that address real-world challenges.
“To address these limitations, we will develop methods to learn the intrinsic properties of the data with probabilistic modeling,” said Avi Pfeffer, chief scientist at Charles River Analytics and principal investigator for PLEDGES.