Adaptation in natural and artificial systems
Adaptation in natural and artificial systems
Some studies in machine learning using the game of checkers
Computers & thought
GPS, a program that simulates human thought
Computers & thought
| Hi-index | 0.00 |
This paper traces the experimental development of a new class of powerful and flexible adaptive plans, called reproductive plans. Adaptive plans are formally presented as search procedures for locating superior devices in an extremely large space. Reproductive adaptive plans operate by treating the search procedure as an evolutionary process of finding the best organism in a certain environment. Devices are represented as strings or chromosomes. At each “generation” or time step a population of devices is tested and each device is copied (rewarded) according to its performance. Then the copies “mate” using a number of genetic-like operators to produce a modified population of devices. In this particular study devices are pattern recognition programs although they could be any set of modifiable procedures. Much of the work is concerned with experimentally testing and improving the general reproductive plan to achieve fast and continuous adaptation.