home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.ai.neural-nets:4309 ont.events:562
- Path: sparky!uunet!ukma!darwin.sura.net!zaphod.mps.ohio-state.edu!rpi!utcsri!csri.toronto.edu!ronr
- Newsgroups: comp.ai.neural-nets,ont.events,ut.dcs.ai
- From: ronr@csri.toronto.edu (Ron Riesenbach)
- Subject: Neural Networks for Industry - 2 Day Tutorial (FREE)
- Message-ID: <1992Nov18.164511.23623@jarvis.csri.toronto.edu>
- Organization: CSRI, University of Toronto
- Date: 18 Nov 92 21:45:12 GMT
- Lines: 159
-
-
- NEURAL NETWORKS FOR INDUSTRY
-
- A 2-Day Turorial by
-
- Dr. Geoffrey Hinton
-
- December 7 & 8, 1992
- Toronto, Ontario, Canada
-
-
- The Annual Neural Networks for Industry tutorial presented by ITRC
- researcher Geoffrey Hinton will now take place at the University of Toronto
- instead of in Montreal. And it's FREE. The tutorial will still run from
- 9.00 am to 5.30 pm on Dec. 7 and 8, 1992. Neural Networks for Industry '92
- is intended for scientists and managers who want an understanding of the
- current state-of-the-art in neural networks and the ability to make informed
- decisions about whether this technology is applicable (or may soon become
- applicable) to specific problems in their area of interest. It focuses on
- the basic principles underlying neural networks, and provides an overview
- of the more significant learning procedures and how these can be used
- effectively in current applications.
-
- Among the topics to be covered are applications of backpropagation, ways to
- improve generalization and speed convergence, competitive learning and radial
- basis functions, elastic matching, unsupervised learning, and a discussion
- of recent developments. To register and obtain details of hotel discounts in
- Toronto, please use the attached form. Space is limited, so please register
- early. This seminar is sponsored by ITRC, PRECARN and CRIM.
-
-
-
- CONTENTS OF DAY 1
-
- 9.00-10.30 INTRODUCTION
- Comparison of computers and brains
- The Least Mean Squares learning procedure
- The perceptron paradigm
- Why adaptive hidden units are needed
- Varieties of learning procedure
- The backpropagation algorithm
-
- 11.00-12.30 SOME APPLICATIONS OF BACKPROPAGATION
- The NetTalk example
- Detecting bombs in suitcases
- Following a road with an autonomous land vehicle
- Recognizing cancerous cells in Pap-smears
- Removing atmospheric distortion from telescopes
- Predicting biochemical processes
- Cross-entropy error functions
- Adaptive interfaces: The Glove-Talk example
- Ways to incorporate expert knowledge
- Forward and inverse models for control
-
- 2.00-3.30 IMPROVING GENERALIZATION AND SPEEDING LEARNING
- Theoretical results on generalization
- Simplicity and generalization
- Varieties of weight decay
- Cross-validation
- Ways of making backpropagation learn faster
- Line search and conjugate gradient
- Cascade-correlation and an application
- Guidelines for applying backpropagation
-
- 4.00-5.30 MORE APPLICATIONS OF BACKPROPAGATION
- Recognizing phonemes in spectrograms
- Recognizing hand-printed digits
- Hidden Markov Models
- Learning the probabilities in HMM's
- Combining neural nets and HMM's
-
-
- CONTENTS OF DAY 2
-
- 9.00-10.30 COMPETITIVE LEARNING AND MAP FORMATION
- Finding clusters with competitive learning
- Soft competitive learning
- The EM algorithm
- Kohonen's method of forming topographic maps
- Methods of forming maps for autonomous robots
-
- 11.00-12.30 COMBINING COMPETITIVE AND SUPERVISED LEARNING
- Radial basis function networks
- Methods of choosing the basis functions
- Some comparisons of learning speed
- Communities of expert networks
- A vowel recognition example
-
- 2.00-3.30 UNSUPERVISED LEARNING AND THE MINIMUM DESCRIPTION LENGTH FRAMEWORK
- The minimum description length framework
- Extracting principal components
- Self-supervised backpropagation
- Cooperative vector quantization
-
- 4.00-5.30 MISCELLANEOUS RECENT DEVELOPMENTS AND GENERAL INFORMATION
- A better way of keeping the weights simple
- The Bayesian approach: Ensembles of networks
- Backpropagation through time and HMM's
- Simulators, workstations, boards, and chips
- Neural network journals and conferences
-
-
-
- ----------------------------------------------------------------------------
-
- Registration Form
-
- Neural Networks for Industry '92
-
- by Dr. Geoffrey Hinton
-
- sponsored by ITRC, Precarn, CRIM
-
- December 7-8, 1992
- University of Toronto
- Galbraith Building
- 35 St. George St., rm. 202
- Toronto, ON
-
- (space is limited, please register early)
-
-
- Name ________________________________________________________
-
- Title ________________________________________________________
-
- Organization________________________________________________________
-
- Address ________________________________________________________
-
- ________________________________________________________
-
- City __________________________ Postal Code___________________
-
- Telephone ____________________________ Fax_________________________
-
- Please fax, mail or email to:
-
- Rosanna Reid, ITRC
- D.L. Pratt Bldg., rm. 286
- 6 King's College Rd.
- University of Toronto
- Toronto, ON M5S 1A1
- fax: 416-978-7207
- email: rosanna@itrchq.itrc.on.ca
-
- Two hotels in close proximity to the University that have good rates are:
- The Delta Chelsea Inn, 33 Gerrard St., W., 416/595-1975; and The Journey's End,
- 280 Bloor St., W., 416/968-0010.
-
- Please mention that you are a University of Toronto visitor so that you may
- obtain corporate/government rates.
-
- ----------------------------------------------------------------------------
-
-
-
-
-
-