An adaptive procedure for controlling undefined linear processes
Kumpati S. Narendra, D. Streeter · IEEE Transactions on Automatic Control · 1964
An adaptive procedure is described and demonstrated which causes a reduction of the mean-square error of a linear stationary random process, without explicit knowledge of the characteristics of the process. The procedure is recursive and, at every step, involves the simultaneous adjustment of an arbitrary number of parameters in a feedback compensator. The adjustment is based on estimates of the error gradient in the space of the adjustable parameters.