Talk:Rescorla-Wagner model

This is a nice article for which I have only two comments.

1. Add a reference to the article "Assessment of the Rescorla–Wagner Model" by Miller, Barnet, and Grahame.

2. In my view, the Rescorla-Wagner model is a "supervised learning" rule. The unconditioned stimulus is the teacher (see Haykin, 1994, page 57).

Hebbian or Widrow-Hoff?

In an earlier version of this article it was stated that the Rescorla-Wagner model is closely related to Hebbian learning. However, the relationship is much closer to the Widrow-Hoff learnng rule, which is explicitly a supervised learning rule. This correspondence was first pointed out by Sutton and Barto in their 1981 paper "Towards a Modern Theory of Adaptive Networks: Expectation and Prediction" in Psychological Review 88, pp. 135-170. This is a very direct correspondence with the US corresponding to the teaching, or target, signal. The paper also discusses simple Hebbian learning and its instability.