Multispectral classification of LANDSAT TM data using a cooperative learning neural network

The authors propose a multilayer multistep backpropagation algorithm which consists of some category extraction networks and a unification network for the land cover classification of remotely sensed data. Each extraction network is used to select the most suitable category. All output patterns of the extraction networks are unified in the unification network. This methodology called a cooperative learning can ensure and accelerate the learning convergence