031407 - Lecture 15: 03/14/2007 Recall: Done with evolution...

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

View Full Document Right Arrow Icon
Lecture 15: 03/14/2007 Recall: Done with evolution strategies. ((1+1)-ES, (μ+λ)-ES, (μ,λ)-ES) Wrap up the EP + ES discussion: 1. EP a. Constant size populations b. Mutation as only source of genetic randomness c. Selection process is deterministic (no “selective randomness”) d. Individuals in the populations were not mutually isomorphic e. Goal: AI-oriented 2. ES a. Also constant-sized populations (μ) b. Early (and bulk of later) work: mutation as only source of genetic randomness; later, some work using recombination of various kinds. c. Deterministic selection process, again d. Isomorphic individuals (vectors in R n ) e. Goal: maximize f(x) over x f. Also, in EP, “environment” is changing (sequence segments you want to match get longer); where as in ES it’s static On to genetic algorithms (GA)-invented in ‘60s, ‘70s by John Holland at U. Michigan-CS. Big picture : A set φ of “candidate problem solutions” and a f: φ ->R called a fitness function , exogenously (ie, from the outside) specified ahead of time. “f(φ 1 )>f(φ 2 )” <-> “φ 1 is a better problem solution than φ 2 Also have a set G of “genotypes” and a surjective
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.

Page1 / 3

031407 - Lecture 15: 03/14/2007 Recall: Done with evolution...

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