CIS 588 Neural Computing Course details. CIS 588 Neural Computing Course basics: Instructor - Iren Valova Tuesday, Thursday 5 - 6:15pm, T 101 1.

CIS 588 Neural ComputingCIS 588 Neural Computing

Course details

CIS 588 Neural ComputingCIS 588 Neural Computing

Course basics:Course basics:

Instructor - Iren ValovaInstructor - Iren Valova

Tuesday, Thursday 5 - 6:15pm, T 101Tuesday, Thursday 5 - 6:15pm, T 101

1 midterm, 1 project, 1 presentation, 3 1 midterm, 1 project, 1 presentation, 3 homeworks, Finalhomeworks, Final

Fundamentals of Neural Networks, Fundamentals of Neural Networks, Laurene Fausett, Prentice Hall, 1995Laurene Fausett, Prentice Hall, 1995

Additional resources are found in the Additional resources are found in the class web site.class web site.

Neural network - what is it?

• 1960s - neural network research preceded the digital computer, but dwindled in 1969 after Minsky and Papert

• 1986 - Rumelhart showed that multilayer perceptron could overcome the limitations described by Minsky

• Rumelhart popularized the notion that there are other viable architectures; by 1989 there were two societies as forum for NN research

• by 1991 people began to realize the significance of computers that could learn new things without having to be explicitly reprogrammed

Learning means behaving better as a result of experience.

Neural network - what can I do with it? Why do I need it?

• with all the attention the NNs have received, there are still only a handful of commercially successful applications; many people have heard about NN, yet few have concept of how to apply them

• NN are exciting because the technology offers the promise of computer system that can dynamically adapt to new situations

• NN only require for the learning algorithm, input signals, and the set that collectively represents the desired behavior, to be specified

• the underlying concept is unlike any of the mainstream approaches and is essential for the successful application of NN

Neural network - Why do I need it?

• computers - biggest bang for the buck, inexpensive, reliable, and fast

• automation problems, NP problems, intractable problems (tasks people do extremely well, but difficult to model)

• brain - limited to operations in milliseconds, but working in parallel, self-organizing

• computers are sequential

Applications of Neural NetworksApplications of Neural NetworksStocks, Commodities, and Futures

Business, Management, and Finance

Medical Applications

Sports Applications

Science

Manufacturing

Pattern Recognition

Stocks, Commodities, and FuturesStocks, Commodities, and Futures

Forecasting Stock Prices– Determines if stock is being underpriced or

overpriced by the market.

Cost Prediction– Predicts the next month's gas price change.

Business, Management, and FinanceBusiness, Management, and Finance

Credit Scoring– Predicts loan application success

Identifying Potential for Misconduct– Predicts misconduct potential based on

employee records.

Finding Gold– Recognizes gold deposits

Medical ApplicationsMedical Applications

Diagnosing Heart Attacks– Recognizes Acute Myocardial Infarction

from enzyme data.

Breast Cancer Cell Analysis– Image analysis ignores benign cells and

classifies malignant cells.

Sports ApplicationsSports Applications

Thoroughbred Horse Racing– Predicts the winning horse in a race.

Dog Racing– Predicts the winning dog in a race.

ScienceScience

Mosquito Identification– Recognizes two species and both sexes of

mosquitoes.

Weather Forecasting– Predicts both the probability and quantity of

rain in a local area.

ManufacturingManufacturing

Welding Quality– Recognizes welds which are most likely to

fail under stress.Computer Chip Manufacturing Quality

– Analyzes chip failures to help improve yields.

Beer Testing– Identifies the organic content of competitors'

beer vapors.

Pattern RecognitionPattern Recognition

Speech Recognition– Voice mail recognition for rotary phone

systems.

Classification of Text– Provides contextual information about text.

ReferenceReference

BrainMaker Neural Network SoftwareURL: www.calsci.com