Skip to content

bradfordw/eaerl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 

Repository files navigation

HOW IT WORKS

1. Randomly generate a population.
2. Create tournament of that population.
3. Run the fitness function on the population of that tournament.
4. Apply the crossover to our two best individuals in the tournament. (80% chance of this happening)
(Think of it as breeding the fittest.  This will yield two children)
5. Perform a mutation (5% chance of mutation) if we hit our mark,
or just return the child as-is.
6. Check our stopping criteria to see if we are done
(we check the two children we just created).
7. Add those children to a new population, which (once it reaches the same
size as our original population) will replace our original (or parent) population.
We define what a "perfect" individual looks like.
Our "perfect" genome is the alphabet in this case.
(Go to step 2 and repeat)

ADJUSTMENTS (how you can change stuff)

-define(CORRECT, "ABCDEFGHIJKLMNOPQRSTUVWXYZ").

By changing CORRECT, you are defining what your desired result should look like. In other words, what you are evolving towards.
You can make this longer, shorter (what fun is that?) and the algorithm will continue to evolve towards it

-define(POPSIZE, 1000). %% Size of our population for each tournament

Changing POPSIZE will just alter the amount of individuals in your population (and how many children to collect before replacing it)

Larger POPSIZEs *should* yield smaller generations before hitting your check constraint and completing.

-define(TOURNAMENT_SIZE, 10).

Changing TOURNAMENT_SIZE will change the number of individuals within a tournament.
You pick (at random) this many (TOURNAMENT_SIZE) individuals from your population.

-define(COUNTER_TABLE,generation_counter).

This is just the name of the ets table that I store a counter in.  In time, there will be more counters (and in theory stats see TODO)

MORE EXPLANATION
See code, there are more explanations in there (via comments that is).

TODO

Print print best of EACH GENERATION (not tournament like it is now).

Break things out of a single script and place them into appropriate modules.
Add tests (yes, they SHOULD have been here first).
Add stats so that we can analyze generally how many generations it takes to get to our check constraint given the population and tourney sizes.
I'm sure much more...

GENERAL

I am no means an expert on EA's this was merely an exercise a friend of mine suggested I try out.
This is the first one I've ever attempted and I've (clearly) got a ways to go with Erlang, so suggest away!

About

An evolutionary algorithm in Erlang

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages