Analysis and design of a practical learning system

A learning machine or system can be defined as a device which modifies its internal parameters in a manner so as to optimize its performance. Learning is one solution to the problems of optimization and stabilization of automatic control systems.
The learning machine discussed in this paper consists of two main sections. The first divides the input phase space into 1,024 regions. The second searches in a trial-and-error manner for the optimum output for each. Before designing the actual system, a preliminary computer investigation was undertaken simulating one complete system to test the feasibility of the above concept.