On Learning in a Time-Varying Environment by Using a Probabilistic Neural Network and the Recursive Least Squares Method.