In this work, a probabilistic model is established for recurrent netwo
rks, The EM (expectation-maximization) algorithm is then applied to de
rive a new fast training algorithm for recurrent networks through mean
-field approximation, This new algorithm converts training a complicat
ed recurrent network into training an array of individual feedforward
neurons, These neurons are then trained via a linear weighted regressi
on algorithm. The training time has been improved by five to 15 times
on benchmark problems.