This is a general question, not related to 1 specific graph…
Each time I run a GD script I get loads of results… Great.
However, the results don’t seem to match what I can achieve as a human by tweaking the sliders manually for a while.
Any tips? Should I just let the GD run for more iterations?
You’re probably not forming problems where generative design is needed. If a problem has an evident solution then it was not necessary in most cases to work that way.
Real generative design is usually complex, in order to provide solutions to complex problems.
Well… I’m testing the stuff on a simplified version (hence me being able to ‘solve’ it)…
But the GD isn’t spitting out as good a solution on the simplified versions.
I seemed to have more success with pure code… Maybe I’m just impatient and should leave the thing running longer.
I’d suggest looking into problems with competing variables and challenging conditions to measure. I most typically use GD for:
Completely agree with @GavinCrump. An example of one of these cases would be helpful as well. It may be something you’re missing in setting up the design space.
Generally I’m running these using Galapagos or Wallacei in tandem with Ladybug tools, I’ll be honest. Company I work at currently, IP etc. but I’ll try to set up a shade optimizer this weekdnd as an example.
I thought ladybug was no more?
Alive and kicking over in Rhino 3d.
Ah. 
Sorry, that was unclear. I meant for @Alien to post an example of one of their graphs that doesn’t seem to give proper results. Although a simple example of a good GD problem would be helpful too.
I got that… But please don’t discourage Gavin from posting summat 
I wanna see 
Hmm, I’m experimenting with changing the scoring…
Maybe a bit of weighting and me running it longer will help.
I think my main issue is impatience. The GD on Dynamo takes ages compared to pure code.
But I do love the graphic output.
If the design space isn’t defined well or there are a lot of input variations it can take a long time to get decent results. Some of these issues can be improved, some can’t.
Also, what do you mean by “pure code”? How and where are you running similar iterations outside of Autodesk GD? Are you still in Dynamo or another platform?
By pure code I mean using Python in Dynamo (so you can use the Dynamo functions too).
It seems a lot faster but obviously you don’t get graphic output.
It’ll run 1000 loops in a second kinda thing whereas the GD seems to take minutes/ hours.
I’ve only done it for simple projects like, optimising areas.
Currently trying to get some shapes to tessellate without over spilling the shape too much.
Have changed the scoring AND set it to loads of runs… Let’s see if it helps.
I have 9 inputs
And just 1 output which is now a total weighted score.
TileFRM (2).dyn (183.2 KB)
Here’s a basic tile thingy…
I have changed the score since then but I’m still not convinced I’m getting great results. I just set it off running again for longer so maybe that’ll help.
Comments welcome
Here’s a simplified example of a study where the results are difficult to determine. I’ve taken a facade and generated a parametric sun shade that can shift in X/Y/Z directions for the front point. I want to optimize the shade profile so it receives sun in winter but prevents sun in summer on the glazing.
To generate the fitness I measure the sun hours received in winter, then take the maximum possible hours from those received in summer to determine overall fitness. An hour blocked in summer is essentially as good as one received in winter on this basis.
We often use Wallacei to access multi-objective outcomes, more close to what you can get in Dynamo/Revit, but added benefit of quick geometry processing times as well as access to the Grasshopper libraries. I’ve attached the graph as well as screenshots as an example. You’ll need to change the ‘dyn’ to ‘gh’ in the extension as the forums don’t allow upload of .gh files. I used Galapagos in this case as it’s much simpler and comes with Grasshopper.
Optimize solstice shading.dyn (61.6 KB)
In above screenshot the summer result is shown, and the same winter result is below. As you can see the solution blocks most sun in summer whilst giving access during winter, so we can effectively find a ‘sweet spot’. Alternatively we could just try to focus on solar radiation (kwh) or other metrics, but sun hours are often a good high level tool to use.
hahahah what grasshopper !!! 
So you’re saying… Don’t use Dynamo?
Not necessarily. Try lots of approaches and find which ones get you the best results that you need.
