Representing a neural network


next up previous
Next: Your own project?
Up: Representing information spatially (AI)
Previous: Parentssiblings, and routes
Back: to main list of student notes

Representing a neural network

Now consider a neural network. You can either think about real biological neurons in the body, or about an artificial neural network. What predicates might you choose to describe such a network? Sketch part of it, and write out a few facts describing your sketch.

Here are some suggestions that might help: Each neuron has one or more inputs from other neurons and one or more outputs. The inputs may be excitatory, or inhibitory. Each neuron has a current activation level. And so on.

Incidentally, Prolog is not a good language for running connectionist networks - its arithmetic is too slow. However, it is an excellent language for describing their structure. Charles Benett, one of our past students, wrote a program called Neuroguide, which contained part of a neuroanatomy atlas, and allowed you to interrogate it and follow around the connections.


next up previous
Next: Your own project?
Up: Representing information spatially (AI)
Previous: Parentssiblings, and routes
Back: to main list of student notes



Jocelyn Paine
Tue Jun 4 17:55:54 BST 1996