Implementing classifications on one bit connection, analog programmable neural network
Emmanuel de Chambost, M. Sonrier · 1989
Summary form only given, as follows. The authors study a binary connection analog programmable neural network (APNN). An integrated 1024/sup 2/ binary connection APNN is assumed to be feasible. A discussion is presented of how any linear classification can be implemented on such a APNN by taking into account the redundancy and the distribution of information among the input vector components. An analog machine, IRENE, has been realized to simulate in a hybrid series-parallel mode an integrated APNN. Applications have been tested successfully on IRENE.>