Aspects of encoder implementation in the context of neural networks

• Main Author: Toh, Cheow Wee.
• Other Authors: K. Arichandran
• Format: Theses and Dissertations
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/19669
• Institution: Nanyang Technological University

There are established results to show that a pattern recognition problem can be handled and trained by a neural network with hidden units. There is also a useful theorem which refers to the type of pattern recognition situations that can be recognised by a perceptron. The perceptron is essentially capable of linear classification. For a given encoder problem, the possibility exists of initially implementing a linear classifier. In the event the linear classifier fails to classify according to a defined convergence criteria, then the implementation with a neural network classifier with one or more hidden units can be considered.