Grasshopper

algorithmic modeling for Rhino

Hi,

First of all, happy new year to everybody!

I just wanted to know if anybody has managed to estimate aproximately the length of galapagos solving time for a definition. Does it have to do with the time Grasshopper takes to solve a definition or with some other factors?


Is it usual that after the first generation done quite quickly, the solver seems to stops and takes very long to make another?

 

Thanks!

 

Cheers

Views: 650

Replies to This Discussion

Hi Miguel,

 

it's very difficult to estimate the remaining time. On linear problems, sure, but it doesn't make a lot of sense to solve linear problems with an evolutionary solver.

 

The time it takes to reach an acceptable solution depends on the following factors (roughly from most important to least important):

 

  1. Time for a single iteration, i.e. the time it takes Grasshopper to update the sliders once.
  2. Navigability of the fitness-landscape, i.e. how easy is it to find a better solution given an existing one.
  3. Fitness target setting. If you demand a very high fitness, it will probably take a lot longer.
  4. Fitness function. Some fitness functions are defined in a way that does not allow a solution at all, others are defined poorly and will result in an unnecessarily long solver time. Creating a good fitness function is not easy. It requires a good understanding of both the problem at hand and the theory in general.
  5. Feedback mode, if the time it takes to refresh the screen is a significant amount of the time it takes to solve a single iteration, you can drastically speed up the process by disabling the preview for every iteration. Use the little speed icons to switch feedback mode in the top right of the Galapagos toolbar.
  6. Chance. Mutations and Coupling contain random elements. Sometimes you get lucky, sometimes you don't.

 

 

Is it usual that after the first generation done quite quickly, the solver seems to stops and takes very long to make another?

 

No. If your population size is enormous, there will be some delay at the end of every generation while the data is parsed and statistics are formulated. But this should not be an issue unless you're dealing with a population size of hundreds or more.

 

I don't know what could be causing this delay in your case. Are you per chance using the "Draw only the best genome in the Rhino viewport" feedback mode? If yes, then the meshing involved with drawing that single frame might account for a delay.

 

--

David Rutten

david@mcneel.com

Poprad, Slovakia

Ok, David, I see it is quite difficult to foresee the solving time without starting galapagos.

Thanks for your explanation

Cheers!

RSS

About

Translate

Search

Videos

  • Add Videos
  • View All

© 2024   Created by Scott Davidson.   Powered by

Badges  |  Report an Issue  |  Terms of Service