tri-j enterprises, inc.

...because good software just matters

  • Increase font size
  • Default font size
  • Decrease font size
Home development genetic algorithms
genetic algorithms

helix fumble

E-mail Print PDF
ok.  we have fiddled around long enough.  we finished the genetic algorithm for the speed meeting application.  the problem is that it does not behave like i thought it would.  so i was reading an article and the crossover rate represents the CHANCE that 2 chromosomes will combine, not that every time i select 2 they will combine and they will cross over at the crossover rate.  does that make any sense?  in other words i was crossing over every 2 that i chose.  now, every time a choose 2 chroms, i will roll the 10 sides dice and see if it 1-7.  if so i will cross over at a random point.  if not, i will discard and just see if they mutate.
Last Updated on Friday, 19 February 2010 20:13
 

genetic progress

E-mail Print PDF

we are continuing development of our speed meeting application.  it is turning out to be a real gem.  this is our first application to integrate a genetic algorithm to minimize duplication in a seating chart.  just last week we got the initial population logic completed.  i am not sure how many chromosomes the app will need in the initial population so we made it variable.  a few days ago we finished the selection process.  the system uses a roulette wheel selection process.  this makes it so that bad chromosomes are kept around to use in child reproduction.  it makes the system converge slower, but all in all it comes to a better solution in the end.  we will keep playing with it and see what happens.  today we got the crossover function written.  we are starting with 70% but we also made this variable to see what happens when we really start to tune the system.  we are using a permutation crossover method because encoding the numbers up to 62, which is how many people we could be seating, was just to nasty and it really doesn't buy us anything.  traditionally encoding takes the system in and out of binary.  needless to say the crossover module is still being tested.  one thing i might start to worry about is performance.  as we develop these modules we are always thinking about ways to keep it fast.  stay tuned cause were going into the mutation logic which should prove challenging.

Last Updated on Friday, 19 February 2010 20:02
 

genetic algorithms

E-mail Print PDF
this is even cooler yet.  we are embarcing on dabbling in genetic algorithms.  we currently are developing an application that does its job without genetics and it does an ok job but with GA's can be some much more.  the only problem is that the application does not use a database connection which might turn out to be a hog of a performer and never come back with a good solution on a personal computer.  we have to do dsome proof of concepts and see what happens.  check back later to see what happens!
Last Updated on Friday, 19 February 2010 20:01
 



search

whos online

We have 13 guests online

stats

OS : Linux m
PHP : 4.4.9
MySQL : 5.5.57-0+deb7u1-log
Time : 16:46
Caching : Disabled
GZIP : Disabled
Members : 70
Content : 55
Web Links : 183
Content View Hits : 74528

Powered by WebRing.