Accession Number : AD1050323

Title :   Promoting Probabilistic Programming System (PPS) Development in Probabilistic Programming for Advancing Machine Learning (PPAML)

Descriptive Note : Technical Report,01 Oct 2013,01 Oct 2017

Corporate Author : Galois, Inc Portland United States

Personal Author(s) : Woldridge,Eric

Full Text :

Report Date : 01 Mar 2018

Pagination or Media Count : 34

Abstract : Machine Learning has demonstrated the potential to transform many areas of science, commerce, and the military. However, creating and maintaining successful machine learning systems is an arduous task that requires a doctoral degree and heroic software engineering efforts. Probabilistic Programming for Advancing Machine Learning (PPAML) by creating probabilistic programming systems and associated solvers-aimed to make existing machine learning applications easier to build and to greatly extend the range of problems that can be successfully solved by machine learning. This effort acted as the voice of the user: (a) exposing the probabilistic programming, machine learning and inference engine performers to a breadth of user scenarios over a wide a variety of domains, (b) evaluated and produced feedback on PPS tools to enable the performer teams to understand user perspectives and spur them to enhance their PPS for future users, and (c) developed a community of users in multiple distinct application areas who are invested in the future developments of PPSs.

Descriptors :   probabilistic models , machine learning , probability distributions , artificial intelligence , programming languages , monte carlo method , algorithms , bayesian networks , engineering , computer science , data set , computer programming

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE