031207 - Lecture 14 Recall Discussing early work on...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Lecture 14: 03/12/2007 Recall: Discussing early work on evolutionary computation. L. Fogel in the 1960’s (evolutionary programming) Goal : evolve “machine intelligence” Focus : evolving finite-state machines for sequence prediction Sequence approx. environment Mealy Machine : eg on board last time—given an initial state and a sequence of input values, machine makes state transitions and puts out output values. For sequence prediction: {input values}={output values}={0,1} To “evaluate” a machine wrt sequence prediction / reconstruction, take initial segment of a sequence as input values; see whether ensuing output values match rest of sequence segment. There are many ways to evaluate / quantify the “degree of matching.” Fogel’s approach: have a sequence to “predict.” Start with an essentially random set of (small-ish) Melay machines; evaluate them on first, say, k symbols in sequence; keep best μ of them to be parents of next generation; create next generation as follows: 1. [genetic Г] Mutate all parents; have them on 2*μ machines 2. Evaluate these on sequence segment + next symbol 3. [selection] Retain the best μ as parents for next generation, go back to (1) *How to mutate? Change initial state Eliminate a sate and randomly reassign its incoming arrows Add a state and randomly assign its outgoing arrows (and get into
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 09/14/2007 for the course ECE 496 taught by Professor Delchamps during the Spring '07 term at Cornell.

Page1 / 4

031207 - Lecture 14 Recall Discussing early work on...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online