Artificial Intelligence

3

Neural Networks

3.1 Introduction

At different points in this book, I will be discussing methods by which machines can learn. In this chapter, I show how the action-selecting computations performed by S-R machines can be learned through exposure to a set of samples of inputs paired with the action that would be appropriate for each input. Although there are many different computational structures that might be used, I concentrate here on networks of TLUs with adjustable weights. Learning is achieved by adjusting the weights in the network until its action-computing performance is acceptable. As mentioned in the last chapter, TLU networks are called neural networks because they model some of the properties of biological neurons. I won’t be speculating ...