🚀 go-pugleaf

ACM SIGART Chicago
June, 1996
News of the Special Interest Group on Artificial Intelligence of the Chicago
Chapter of the Association for Computing Machinery

NEWS AND NOTES

Please note upcoming meeting dates on your calendar now so you wont miss a
single meeting:
        June 6, 1996 - Probabilistic Neural Networks, John Garton
        September 5, 1996 - Dale Reed will present a talk entitiled Borrowing
Others' Ears: A Perceptual Assistant for Sound Equalization
        December 5, 1996 - To be announced

This months talk centers on a lesser known type of Neural Network - the
Probabilistic Neural Network.  Also, we will be back at our usual IIT Kent Law
School meeting location.

Special thanks to Michael Liebman and Steven Lytinen for their participation in
the panel discussion on Expert Systems at the March meeting.

June Meeting Announcement

Probabilistic Neural Networks
by John Garton

Abstract:

Say the words Neural Network and everyone automatically thinks of the
traditional backpropagation network. This architecture represents the best known
of the multiple-layer feedforward networks (MLFNs). It was popularized by the
Parallel Distributed Processing books of Rumelhart and McClelland in the
early-mid 1980s. However, there exists a diverse world of neural networks
beyond backprop. The Probabilistic Neural Network, or PNN, is based on work
began by Donald Specht in the late 1980s.  It has some virtues not found in
traditional backprop networks. The PNN has the ability to train very quickly.
It is often several orders of magnitude faster in training than a backprop
network.  Also, one of the most interesting facets of the PNN is its ability to
provide mathematically sound confidence levels for its results.  It also has the
ability to make good predictions when provided with sample data from sparse
domains.  In addition to the theory of PNNs, this talk will examine an actual
C++ implementation of the network.  Some discussion will also center on the
effects of an object-oriented vs. procedural implementation of this type of
network. Parts of this talk will draw heavily on an excellent book by Timothy
Masters, Advanced Algorithms for Neural Networks.

Speaker:

John Garton is president of Byte Knights, Inc., a Chicago-based consulting firm
specializing in Artificial Intelligence and Multimedia Systems. The company is
currently producing an entertainment program that will track and predict lottery
numbers using a PNN. John has an MS in Artificial Intelligence from IIT, and a
BS in Computer Science from CCNY. John also serves as the volunteer chairperson
of the SIGART group. He has previously spoken to the SIGART group in 1992 on the
topic An Introduction to Neural Networks.


Date: Thursday, 6/6/96
Time: 5:30 p.m. Refreshments & Snacks; 6:30 p.m. Speaker
Place: IITs Chicago Kent Law School
          565 West Adams St., Chicago
Cost: $8.00 Members; $10.00 Non-Members; $4.00 Students
Reservations: Call the ACM Faxback Line, (800) 615-5679, Menu option 4, and
leave a message with your name. Please DO try to make a reservation, it really
helps us plan the meeting space.
Information: Call John Garton, (312) 327-9250.

Travel Directions to SIGART Meeting
Via Metra:      Use Union Street Station.  Walk West on Adams St.
Via CTA:        Take Ravenswood, Evanston, or Midway trains to Adams St.  Walk
                  West on Adams St.
Via Bus:        41, 60, 125, 151, 156, 157 to Union Station.  Walk West on Adams
                  St.
Via Car:        Kennedy/Dan Ryan to Monroe exit.  Go East to Clinton, South to
                  Adams.  On street parking is limited.  
                  There are several parking lots in the area.
Walking:        Walk West from the Loop on Adams.

Chicago Chapter ACM
PO Box 2381
Chicago, IL  60690

--