Page 29 - Artificial Intelligence in the Age of Neural Networks and Brain Computing
P. 29
16 CHAPTER 1 Nature’s Learning Rule: The Hebbian-LMS Algorithm
FIGURE 1.11
A learning experiment with a single neuron. (A) Initial responses. (B) Responses after 100
iterations. (C) Responses after 2000 iterations. (D) Responses after 5000 iterations. (E) A
learning curve.
A learning curve, mean square error as a function of the number of iterations, is
shown in Fig. 1.11E. When using a supervised LMS algorithm, the learning curve is
known to be a sum of exponential components [7]. With unsupervised LMS, the the-
ory has not yet been developed. The nature of this learning curve and the speed of
convergence have only been studied empirically.
