Multispectral classification of LANDSAT TM data using a cooperativelearning neural network

摘要

The authors propose a multilayer multistep backpropagationalgorithm which consists of some category extraction networks and aunification network for the land cover classification of remotely senseddata. Each extraction network is used to select the most suitablecategory. All output patterns of the extraction networks are unified inthe unification network. This methodology called a cooperative learningcan ensure and accelerate the learning convergence