Accession Number:

ADP007151

Title:

Simulation Experiments for Neural Network Learning,

Descriptive Note:

Corporate Author:

BOEING COMPUTER SERVICES CO SEATTLE WA

Personal Author(s):

• Newman, David S.

Report Date:

1992-01-01

Pagination or Media Count:

4.0

Abstract:

This paper investigates approaches to the design of simulation experiments for training neural networks which are to be used as classifiers. Hierarchical clustering applied to the ART1 and ART2 ART Adaptive Resonance Theory neural network architectures developed by Carpenter and Grossberg 20,21 is the basis for the approach. A series of experiments based on this approach will test the performance of ART1 and ART2 as pattern classifiers against a variety of real and artificial data sets. The issues to be investigated in these experiments include the sensitivity of performance to a variety of network parameters, pattern characteristics, and pattern presentation disciplines. A background is provided for those unfamiliar with neural networks in general, and with Grossbergs approach in particular.

Descriptors:

• *CLUSTERING
• *NEURAL NETS
• APPROACH
• BACKGROUND
• NETWORKS
• PAPER
• PARAMETERS
• PATTERNS
• RESONANCE
• SENSITIVITY
• SIMULATION
• TEST AND EVALUATION
• TRAINING

Subject Categories:

• Biology

Distribution Statement:

APPROVED FOR PUBLIC RELEASE