Implementation of Fast Orthogonal Search algorithm
FOS is used to identify an ARMA model. reference: Applications of fast orthogonal search: Time-series analysis and resolution of signals in noise
usage:
1.model = FOSModel(max_delay_in_input, max_delay_in_output, max_order, max_m, mse_reduction_threshold), where max_m is the max number of terms, mse_reduction_threshold is the early-stop criteria
2.model.fit(x, y), x (input) and y (output) are 1-dimentional series
3.model.predict(x, y), all x points and a few initial points of y, predicted y series will be returned