Accession Number:

ADA279073

Title:

Backpropagation and EEG Data

Descriptive Note:

Final rept. Aug 1989-Aug 1990

Corporate Author:

ARMSTRONG LAB WRIGHT-PATTERSON AFB OH

Personal Author(s):

• Morton, Paul E.
• Wilson, Glenn F.

Report Date:

1988-10-01

Pagination or Media Count:

10.0

Abstract:

The development of neural networks has pursued a myriad of different courses reflecting the interests of a large number of researchers from highly varied backgrounds. This paper would like to focus on one point of this many faceted gem, as Stephen Grossberg described the field. The point of focus will be to address some of the practical results of applying a backpropagation trained net to raw electroencephalogram EEG data. Much important work on more efficient training rules has been done however, equally critical is consideration of the information content of the data, the net size, number of hidden nodes and order of training data. This paper explores some of the training issues raised by applying backpropagation to this very complex data.

Descriptors:

• *ELECTROENCEPHALOGRAPHY
• NEURAL NETS
• CLASSIFICATION
• RESPONSE(BIOLOGY)
• MEDICAL RESEARCH

Subject Categories:

• Psychology
• Medicine and Medical Research

Distribution Statement:

APPROVED FOR PUBLIC RELEASE