Deep Learning Explained

Co-sponsored by: ASME and SME Westchester Chapters

Abstract: Starting with a few seminal papers in the latter half of the 1980s, the training of multi-layer neural networks was made practical and these networks were soon used for applications in advanced nonlinear pattern recognition, control systems, and many other problems.  Although neural network applications mushroomed in the 1990s, complex architectures were still not very feasible due to the tremendous amount of computation needed for the training of deep neural networks.  The advancements in Graphical Processing Units (GPUs), in the form of massive parallelization capabilities at relatively low costs, created the grounds for another revolution in the field.  The new resurgence of these networks was coined deep learning after the much more complex and deeper architectures.  GPUs made deep neural networks a lot more accessible to the general scientific population and many more applications which used these techniques.  Examples are speech recognition, speech synthesis, face recognition, financial applications, control systems, digital assistants, expert systems, and many more.  The talk will provide an explanation of these networks and will discuss some of the architectures as well as a demonstration of speech recognition using deep neural networks.  An overview of how the learning happens in these networks is also presented.

Speaker(s): Dr. Homayoon Beigi,

Agenda:

6:30 pm – Refreshments and networking

7:00 pm – Introductions and opening remarks

7:10 pm – Presentation starts

7:55 pm – Questions and Answers

8:00 pm – Presentation ends and closing remarks

Location:
Room: Room 104 E/W
Bldg: School of Health Sciences and Practice (SHSP)
New York Medical College
30 Plaza West off Sunshine Cottage Road
Valhalla, New York
10595