# Real value encoding in Genetic Algorithm

Page 1 of 1

## 3 Replies - 8525 Views - Last Post: 30 July 2012 - 09:44 PM

### #1 Saimmehrish

Reputation: 0
• Posts: 2
• Joined: 28-June 12

# Real value encoding in Genetic Algorithm

Posted 28 June 2012 - 06:22 AM

I am working on Genetic Algorithm, there is no issue in coding but if there is any expert of Genetic Algorithm then i would like to ask few technical problems. by the way i am doing real value encoding in my optimization research work, so if some body have already worked on that kindly let me know. there is issue in handling the algorithmic flow.my email id is ***REMOVED EMAIL ADDRESS***, will b waiting for warm response.

This post has been edited by JackOfAllTrades: 28 June 2012 - 06:53 AM

Is This A Good Question/Topic? 0

## Replies To: Real value encoding in Genetic Algorithm

• Saucy!

Reputation: 6246
• Posts: 24,014
• Joined: 23-August 08

## Re: Real value encoding in Genetic Algorithm

Posted 28 June 2012 - 06:55 AM

### #3 Saimmehrish

Reputation: 0
• Posts: 2
• Joined: 28-June 12

## Re: Real value encoding in Genetic Algorithm

Posted 28 June 2012 - 12:05 PM

Look when we are working with genetic algorithm, we do the following steps:
1. We generate initial random population of size N
2. we then select two random population by using Roulette Wheel or any other approach.
3. then we apply crossover operator
4. apply mutation operator.

Now the question is that:
What if the mutated off springs are less fit then initially selected two parents??

and if these mutation off springs are improved then for next iteration we need to put these two population in initially selected N population and then again select two parents

or directly apply crossover and mutation again and again on mutated off springs?

### #4 friday13

Reputation: 1
• Posts: 5
• Joined: 24-July 12

## Re: Real value encoding in Genetic Algorithm

Posted 30 July 2012 - 09:44 PM

By starting off with a random population, selecting certain "fit" members to survive, and other members (fit or not) to mutate or perform crossover, you create a second generation. Do the same, create a rhird generation...and so on. You are using several means of selection to create the next generation. The offspring of parents selected are created by crossover. Other offspring are created by mutation. You may also select certain fit members of the population to carry over to the next generation. The idea is to create some change from generation to generation; a little diversity is a good thing.

I don't really think of it as keeping most of the first generation, but more about creating a second one made up by those different methods from the first. By using both fitness and randomness, each generation has a chance to improve.

You are right - some of the offspring may end up being less fit, but by introducing diversity and performing the algorithm over many generations, you should make progress in solving your problem. It may take some experimentation to come up with the right proportions to perturb to get the gains you are seeking.