Summary of the characteristics of two classes of predictive model.

The equations are typical instantiations of each model class—for the statistical endpoints distribution model, a temporal difference learning rule in which the value of an item on iteration (Vt+1) is equal to the value on iteration t, plus some proportion of the prediction error (δ) times learning rate (α). Dynamic forward models would typically be captured by a set of differential equations, where the rate of change of some parameters (Θ), such as position, is a function of time, f(t).