CODEWITHSUNDEEP
genetic-algorithm
Zach
Zach

Reputation: 1321

How can I represent a number as a chromosome in a genetic algorithm?

I want to use a genetic algorithm to solve a simple system of two linear equations with two variables. This is mainly to help me get a better understanding of how they work.

Everything seems pretty simple, but I am unsure how to encode possible solutions in the chromosomes for this problem.

I will have two variables which I want to encode in a chromosome to represent a solution. If each variable is can be represented an 8-bit number, would I make a 16-bit binary encoded chromosome (A string of 1's and 0's).

I am just not quite sure how that would work. If two parents are selected for breeding, how would randomly selecting genes from the binary string result in a possibly better solution? This is why I don't think a binary string would work, so any answers would be greatly appreciated!

Upvotes: 2

Views: 1735

Answers (1)

Andreas
Andreas

Reputation: 6475

Why not use the numbers as numbers? You don't have to use binary encoding in a GA. There are mutation and crossover operators working well for real-valued encodings. As you say it's a learning example.. I would recommend you try both approaches, the real-valued encoding should be much quicker to converge.

For binary encoding I would use Single Point Crossover and Bit flip Mutation. For real-valued encoding I would use Blend-Alpha-Beta Crossover (BLX-a-b) or Simulated Binary Crossover (SBX) and Normal Distributed Mutation. You can try some of these and many more operators on the SingleObjectiveTestFunctions in HeuristicLab.

Upvotes: 6

Related Questions