← A generalized LSTM-like training algorithm for …