Deep Learning and Grammatical Inference

Time and Date: 3:00 PM, Monday, October 22, 2018

Place: 235 Weir Hall

Speaker: Dr. C. Lee Giles

 

Abstract: Deep learning (DL) is a rather new machine learning model strongly based on neural networks. It became popular primarily for performance reasons, especially when it outperformed other machine learning methods, first in the areas of computer vision, then famously in machine translation, all in this decade. Since then DL has continued to perform well on many large data problems and in related competitions and applications seem to proliferating.

DL architectures can loosely be divided into two classes, stateful and stateless, or respectively recurrent or dynamic networks vs feedforward networks. In practice hybrid systems combining both often exist. There has been a recent rediscovery of neural networks and memory structures, primarily because of the computational power such models offer. We discuss some of these memory model DLs. And we see how DLs can represent and learn state machines. Surprisingly state machines can be extracted and interpreted from these DL.

 

Speaker's Bio: Dr. C. Lee Giles is the David Reese Professor at the College of Information Sciences and Technology at the Pennsylvania State University, University Park, PA. He is also graduate college Professor of Computer Science and Engineering, courtesy Professor of Supply Chain and Information Systems, and Director of the Intelligent Systems Research Laboratory.  He recently became a Teaching and Learning Technology Fellow and the Interim Associate Dean of Research for IST.