The particular pattern recognition device considered herea linear threshold deviceprocesses the input signals di(t) by first multiplying each one by some weight wi and then summing them to yield. When this sum exceeds a given fixed threshold K the input configuration will be said to be a member of the set C+, otherwise a member of the set C-. (It should be clear that and that is empty, so that the linear threshold device partitions C into two classes.) More precisely C+ is supposed to be an approximation to C1,so that when the sum exceeds the fixed threshold K,the device indicates (rightly or wrongly) that the input configuration is an instance of the pattern. The object of the adaptive plan, then, is to discover as rapidly as possible a set of
. When this sum exceeds a given fixed threshold K the input configuration will be said to be a member of the set C+, otherwise a member of the set C-. (It should be clear that 
and that 
is empty, so that the linear threshold device partitions C into two classes.) More precisely C+ is supposed to be an approximation to C1,so that when the sum exceeds the fixed threshold K,the device indicates (rightly or wrongly) that the input configuration is an instance of the pattern. The object of the adaptive plan, then, is to discover as rapidly as possible a set of