I was wondering if anyone new how generative design outcomes will be affected by differing scales of numbering. How are they weighted?
Would a 1 to 100 numbering output have more, less, or no impact when also comparing an input of 1 to 1000. Should my outputs be converted to % ranges or does it matter?
The question came up because I am wondering how to have Generative Design place more concern on one variable over the other while the other is important but maybe secondary. This is accounted for by the user at the end but I wanted to make sure the differing scales of numbering illustrated below do not cause a weighting issue.
Thanks all
The size of the difference will have some impact on the mutation phase of the genetic algorithm. It’s not much, but it’s some. As a result I find it’s often best to have inputs set to be a percentage value rather than a sum or count (ie; what % of trees do I cut; what % of units are on the less expensive soil condition; what % fo the overall land is utilized by units; etc).
Another alternative for weighting values is to use the ‘stronger’ value as a penalty for all values. I often use “does the building fit on site’ to illustrate this when utilizing unbound UV parameters to place elements - if the area of the building intersected with the area of the site isn’t the same as the area of the building, multiply all evaluation values by 100, otherwise multiply by 1. This ensures that the computer always finds solutions which fit on site, otherwise EVERY value is 100 units too large.
