SkyKnit: When knitters teamed up with a neural network

Fun, but if you want to see some seriously interesting modern work in knitting check this out: https://www.ri.cmu.edu/event/on-demand-machine-knitting/

Wow, it is amazing to see someone turn such a seemingly mundane subject into an interesting talk!

That was awesome. I really like the presenter's style, too.

I'm kind of curious about the person whos mental construct crashed. How long did it take before her mind was able to interpret the math again? Was it a slow boot up or was it instant?

Is there a reliable training method to minimize the need for "debugging" in data sets like this?

I would think there is a way to tell the system some basic rules, so its not completely unsupervised.

Like: Here is a list of all stitches (so it doesn't create its own), and the number of stitches must be correct for dimensions, or something like that. Then you validate all the input data against that, and move on from there.

Its a bit more complicated than that, since the only constraints is the yarn and physics. And there's infinite possibilities of stitching things together in intricate manners. Then deciding whether something is nice/wanted, is in the eye of the beholder :)