Classification of Multi-Spectral Pixels by the Binary Diamond Neural Network

Part of Advances in Neural Information Processing Systems 6 (NIPS 1993)

Bibtex Metadata Paper

Authors

Yehuda Salu

Abstract

A new neural network, the Binary Diamond, is presented and its use as a classifier is demonstrated and evaluated. The network is of the feed-forward type. It learns from examples in the 'one shot' mode, and recruits new neurons as needed. It was tested on the problem of pixel classification, and performed well. Possible applications of the network in associative memories are outlined.

1

INTRODUCTION: CLASSIFICATION BY CLUES

Classification is a process by which an item is assigned to a class. Classification is widely used in the animal kingdom. Identifying an item as food is classification. Assigning words to objects, actions, feelings, and situations is classification. The purpose of this work is to introduce a new neural network, the Binary Diamond, which can be used as a general purpose classification tool. The design and operational mode of the Binary Diamond are influenced by observations of the underlying mechanisms that take place in human classification processes.

An item to be classified consists of basic features. Any arbitrary combination of basic features will be called a clue. Generally, an item will consist of many clues. Clues are related not only to the items which contain them, but also to the classes. Each class, that resides in the memory, has a list of clues which are associated with it. These clues