072 Logical Models of Neural Networks

I want to start by giving a very sketchy account of neurophysiology- merely sufficient as a basis for our first mathematical model. We may regard the human nervous system as a three-stage system as shown in Figure 2.1.* Our fundamental hypothesis in setting up our model is that all the functioning of the nervous system relevant to our study is mediated solely by the passage of electrical impulses by cells we call neurons. Actually, the human brain contains more glial cells than it contains neurons, but it is neurophysiological orthodoxy to believe that these glial cells served only to support and nourish the neurons–functions irrelevant to our study. Throughout this book, we shall ignore such posited glial functions. We shall also ignore such modes of neural interaction as continuously variable potentials and transmission of hormones. In setting up our possible mechanisms, neural impulses will fully suffice-future developments will, of course, require the ascription of far greater importance to the other neural functions and perhaps to the glia. In the light of our fundamental hypothesis, then, we shall simply view the nervous system proper as a vast network of neurons, arranged in elaborate structures with extremely complex interconnections. This network receives inputs from a vast number of receptors: the rods and cones of the eyes; the pain, touch, hot, and cold receptors of the skin; the stretch receptors of muscles; and so on; all converting stimuli from the body or the external world into patterns of electrical impulses that convey information into the network. These interact with the enormously complicated patterns already traveling through the neural net (there are estimated to be more than 10¹⁰ neurons in the neural net which is the human brain) and result in the emission of impulses that control the effectors, such as our muscles and glands, to give our responses. Thus we have our three-stage system: receptors, neural net, and effectors.