Grasshopper

algorithmic modeling for Rhino

Here's a weird one: my file crashed at the end of the day yesterday, but well after it had been saved with a WallaceiX component holding 50,000 solutions.

I baked the Pareto front (yesterday, on the left in the attached images), and it essentially looked as roughly what I would have expected. However, when I open up WallaceiX and re-export the Pareto front today, I get the results on the right.

Am I missing something silly/obvious here? As you can (sort of) see from the baked text, the fitness values are the same, but the geometries are obviously quite off.

Thanks,
Magnus

Views: 58

Attachments:

Replies to This Discussion

Ill have a look at your definition once I'm infront of my machine - but quickly glancing at your images - you are running a simulation with 32 fitness objectives, is that intentional?

Hi again Mohammed, 

Nope - so we're squarely back in "silly mistakes" country (sorry about that). 

There should be four objectives. I'll begin going over this on my (slow) machine now; if you've got a time to give it a quick glance later on and give me some pointers, that would be truly appreciated.

Many thanks,
Magnus

Hmmm.

Here are the screenshots from before and after running the simulation - looks like four objectives, so something else must be messing this up along the way...

The plot thickens.

/m 

Attachments:

Hi Magnus,

I think I found the problem (or parts of its atleast) - you are grafting your genepools (which in theory is fine), but because you are grafting the genepool component itself, when you connect it into Wallacei, it receives the genes in different branches - Wallacei doesnt like fitness objectives or genes to be in different branches, it likes them to be in a single flattened list. I ran a quick test where I flattened your genes, and it seems to work fine - I believe this may also be contributing to your pcp time problem but cant tell for sure without re-running the simulation again (44 hours of which I dont currently have!)

Also, your objectives are in different branches (they are in different tree paths).

So, please do the following (essentially before running any simulation):

1. Dont graft the genepool component, rather, use a graft component instead, so that when you connect the genepools to Wallacei, they are in a single flattened list.

2. Always flatten the number component holding your fitness objectives before connecting them to Wallacei.

Let me know if the above fixes the problem.

Cheers Mohammed - many thanks for taking your time with this, and for the very clear and straightforward instructions. I'll add the graft components, and no fitness objectives components will ever be anything but flattened from now on.

Come Friday I'll have another 44 hours (or so), and so I'll re-run this and see what difference it makes.

All best,
Magnus 

RSS

About

Translate

Search

Photos

  • Add Photos
  • View All

© 2019   Created by Scott Davidson.   Powered by

Badges  |  Report an Issue  |  Terms of Service