Probabilistic Programming for Advancing Machine Learning

Kathleen Fisher, , Jun 03, 2013
Commentary by Stephen Downes
files/images/DARPA_Program_structure.JPG

Take the time to give this presentation a good listen (the video is in user-hostile Quicktime, so give it lots of time to load). It outlines requirements for DARPA's latest round of research funding on probabilistic programming and machine learning. What's important is not the funding opportunities (which, believe me, none of us can qualify for) but the description of where (and how) analytics and chaotic data processing is headed. The days where a computer program was just a set of instructions has long since passed. The future is computing based on models, queries and fact sets (there's a really good example predicing outcomes of games of tug of war, which is based on the TrueSkill ranking system used by xBox; anotherexample looks at recommendation systems such as Matchbox's Large-Scale Baysean Recommendations). The impact of this changing paradigm drives deep into what we think are the different skills in learning and design - as illustrated, we have SMEs, who pose questions, problem-language and model-language programming experts, machine learning design experts, and compilers and cloud storage experts. All very different, all vital.

Views: 0 today, 123 total (since January 1, 2017).[Direct Link]
Creative Commons License. gRSShopper

Copyright 2015 Stephen Downes ~ Contact: stephen@downes.ca
This page generated by gRSShopper.
Last Updated: Aug 20, 2017 05:51 a.m.