Evolution of a Tri-State Frequency Discriminator for the Source Identification Problem using Genetic Programming
John R. Koza, Jason D. Lohn, F. Dunlap, Martin A. Keane, Martin A. Keane, David André · 2004
Automated synthesis of analog electronic circuits is recognized as a difficult problem. Genetic programming was used to evolve both the topology and the sizing (numerical values) for each component of a circuit that can perform source identification by correctly classify an incoming signal into categories. 1. Introduction The problem of source identification involves correctly classifying an incoming signal into a category that identifies the source of the signal. The problem is difficult because it requires discovery of the features that distinguish the sources and because no two real-world signals from the same source are ever exactly the same. This paper considers the problem of evolving the design for an analog electrical circuit that can solve the problem of source identification for incoming signals coming from sources that each emit signals lying in a certain limited range of frequencies. Each incoming signal is to be identified as coming from the first source, the s...