Grid of generated solutions

[johnmirts]

Hello John,
I am writing a "BiomorpherReaderAdvanced" component which shall return all the generated solutions as a list.
For this attempt I have used the provided file "Vases.gh" and thus the component only considers Rhino.Geometry.Mesh objects.
I haven't gone that far, as I am a bit stuck. Would you be interested in giving me a helping hand?

Upon returning the list of solutions, I am planning to plot everything on Rhino canvas arranged at a grid with the help of an additional component, which is almost ready. I could share it via a "pull request".

Best,
Ioannis

BiomorpherReaderAdvanced.zip

[johnmirts]

I suspect that all the magic is happening or better, should be happening here:

canvas.Document.Enabled = false;
SetSliders(genes, theSliders, theGenePools);
if (DA.GetData(2, ref mesh)) {
    allMeshes.Add(mesh);
}
else {
    // add an error message
}
canvas.Document.Enabled = true;
canvas.Document.ExpireSolution();

Indeed, the list of Meshes has the length of the generations*designs, but they are all identical and the components on the GH canvas are continuously updating.....

[johnharding]

Hi Ioannis. Sorry I completely missed this, didn't have the github notification settings on!
Are you still working on this? I was thinking of implementing something similar, however it ended up being quite memory intensive to store the actual geometry (hence the reader component instead adjusting the sliders and only storing genes (parameter states). You can also use a record component to store meshes each time Biomorpher displays per iteration, but perhaps your intentions mean this is not possible.
Let me know if you still need help.
John.

[johnmirts]

Hi John. I have stopped working on it. I figured out the same thing (it is memory intensive). So finally I ended up using the data recorder as you suggest. Recently an other thing that concerns me is phenotype-based clustering rather than genotype-based. I will probably dive into it in 2 months. Do you have any suggestions/references/thoughts?
Best and thanks for this amazing tool,
Ioannis

[johnharding]

Ah ok, great. Yep things can spiral out of control with large populations of geometry!
Yes I've thought about implementing it for performance measures, and then comparing with the gene clustering. This gives an overall feel of how 'direct' the mapping between genotype and phenotype is also in the definition/analysis.
I guess here papers around Evolutionary Developmental Biology (Evo-Devo) would be worth looking at, as GAs work well when the encoding is neither completely direct (for example linear relationship between slider and property) or indirect (for example, slider controls random seed), but somewhere in between. I believe it's also called an 'evolutionary embryogeny' or something like that!
Let me know how you get on.