Created
October 17, 2011 02:05
-
-
Save lspector/1291789 to your computer and use it in GitHub Desktop.
A simple binary genetic algorithm in Clojure, to demonstrate one way to write an evolutionary loop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
(ns evolvesum) ;; Lee Spector (lspector@hampshire.edu) 20111009 | |
;; We evolve a vector of 100 zeros and ones that sums to a particular number. | |
;; An individual is a vector of 100 random bits. | |
(defn new-individual | |
[] | |
(vec (repeatedly 100 #(rand-int 2)))) | |
;; An individual is mutated by possibly flipping a random bit. | |
(defn mutate | |
[individual] | |
(assoc individual (rand-int 100) (rand-int 2))) | |
;; The error of an individual is the difference between the sum of the bits | |
;; and the goal, which we hardcode here. | |
(defn error | |
[individual] | |
(Math/abs (- (reduce + individual) 73))) | |
;; An individual is better than another if it has lower error. | |
(defn better | |
[i1 i2] | |
(< (error i1) (error i2))) | |
;; We evolve a solution by starting with a random population and repeatedly | |
;; sorting, checking for a solution, and producing a new population. | |
;; We produce the new population by selecting and mutating the better half | |
;; of the current population. | |
(defn evolve | |
[popsize] | |
(loop [population (sort better (repeatedly popsize new-individual))] | |
(let [best (first population)] | |
(println "Best error:" (error best)) | |
(if (zero? (error best)) | |
(println "Success:" best) | |
(let [better-half (take (int (/ popsize 2)) population)] | |
(recur | |
(sort better (map mutate | |
(concat better-half better-half))))))))) | |
;; Run it with a population of 100: | |
(evolve 100) | |
;; Exercises: | |
;; - Create variables or parameters for the various hardcoded values. | |
;; - Avoid recomputing errors by making individuals pairs of [error genome]. | |
;; - Print more information about the population each generation. | |
;; - Select parents via tournaments. | |
;; - Add crossover. | |
;; - Use a more interesting genome representation, for a more interesting | |
;; problem. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment