About the logic of Generative Design

I would like to know more about the optimize of Generative Design.
I have read the information at the URL below, but is there any other information on “optimize” in detail?

I would like to get information that will help me solve the following questions

  • When I select “optimize” on the study screen of Generative Design, is there any rule for the results that are displayed from the upper left corner of the screen?


There are rules but I am not sure what they are. I am sure that the specifics aren’t really important - you and I can think of them as the top performers for a given metric.

Say you have a study which optimizes metrics A and B. If you watch the results in Generative Design carefully you’ll see a handful of results on the first run, all of which perform better for metric A or B, but likely none of which perform well in both. On the second run you might see options which perform better for A, but B might not have improved, or may have even gotten worse (depending on how your metrics are setup). On the third run A or B might get better, and again on the 4th, 5th… eventually getting to the final run where ideally A and B have both found their ideal state.

The number of options displayed will also change over time. If there is 1 really good output for A at any point, but 10 ok ones, you might see all 11 results. But once A finds 4 really good options the remaining 11 will drop as they are too far removed from the bulk of the dataset. This also occurs for B at the same time, as well as the combination of A and B. Effectively there is a vector of each result, and anything inside the upper range of what the machine as found possible is displayed in the “hall of fame” which are the options exposed to us as users.

1 Like

FYI, you can find a report of the Hall of Fame as well has the full history results in %APPDATA%/GenerativeDesign. That might give you a better idea of what’s happening between each generation.