Modeling with Neural Networks: Principles and Model Design Methodology - Neural Networks: Methodology and Applications - page 159

Fig. 2.39. NARMAX assumption

Fig. 2.40. NARMAX model

Table 2.1. Noise assumptions and their consequences on the training of input-
output models

Usual name in Equivalent in Recommended

Assumption nonlinear linear modeling Training operation

State noise ARX Directed One-step-ahead

predictor

Output noise Output error Semidirected Simulator

State noise and NARMAX ARMAX Semidirected One-step-ahead

output noise predictor