Learning in Perceptual Control Networks

NTT Systems worked with Dr. Martin Taylor at DCIEM on a project to build perceptual control networks.

The project (not yet past the proof-of-concept stage) is to build hierarchical perceptual control networks for the predictive tracking of the output of a toy grammar. The grammar was meant to reflect some of the basic properties of human speech. The goal was for the system to reorganize and learn new strategies for predicting the grammar's output, mimicking the way humans predict as they track speech during passive listening. Tools were built in the Prograph language, on top of a NTT Systems' MVC (Model-View-Controller) system, for experimenting with control systems. Only initial results were obtained.

Other Neural Net projects Other Prograph projects
Please visit our home page: NTT Systems Inc.