|Intelligent Systems And Their Societies||Walter Fritz|
Feedback, reinforcement, is a very important mechanism that increases the value of some response rules and decreases the value of others. Feedback is provided to the brain from two very different sources:
Traditionally, there are two, very unequal types of reinforcement: negative and positive. Negative reinforcement makes less probable the use of a non-optimal response rule. However, it does not show which response rule to use in the future instead. If, on the other hand, a good response receives positive reinforcement, the IS will both increase the values of the response rule it used, and, in the next sleep period, create a more general response rule of high value that it is then likely to use in the future.
Thus, we can see that this is a classic case where the "carrots" of approval that characterize positive feedback have a different and much more useful effect in the learning cycles of an IS than does only the "stick" of negative reinforcement.
For continuos reading, like a book - continue
Jump to the top of this document /Details of IS /Intelligent Systems /E-Book Contents.
Copyright © New Horizons Press