Blog: Machines That Learn Better

Machines That Learn Better
MIT News (05/18/10) Hardesty, Larry

Massachusetts Institute of Technology (MIT) researchers have developed Church, a probabilistic programming language designed to cut the time it takes to build a machine-learning system to a matter of hours instead of months. Church is based on an inference algorithm, which instructs a machine-learning system on how to draw conclusions from the data presented to it. The algorithms currently used in probabilistic programming are designed to handle discrete data but struggle with continuous data. However, at the recent International Conference on Artificial Intelligence and Statistics, MIT student Daniel Roy presented a paper in which he and MIT instructor Cameron Freer describe an inference algorithm that can handle large classes of problems involving continuous data. Their work could be especially useful for artificial intelligence systems whose future behavior is dependent on their past behavior, says Rutgers University computer scientist Chung-chieh Shan.

View Full Article