I was just reading about JSD the other day after reading about KL divergence...seems like a nifty measurement device for things like sim-to-real evaluations in robots (the reason I was going down this rabbit hole.)
I think the appeal over raw KL is that JSD behaves a bit nicer when the simulated and real distributions don't perfectly overlap...which is basically always true in the real world!
I did some river/lake sailing as a kid on the East Coast but now the urge is calling to me! I remember the "righting the boat" test being the scariest/most fun part of the experience -- super glad I went through that and feel confident on a small boat.
Now...I used to remember all the knots we learned but that memory is mostly gone
mold making is also pretty complicated -- anything in the 1,000-1M parts produced will _probably_ be an aluminum mold (cheaper than steel) but they're still heavy and large to keep around.
I haven't met any injection molding shops in the US that do a huge amount of specialty parts like toys. The industry tries to get as many medical device jobs as possible.
This scratches a corner of my brain I didn't know I had. It's the start of the month and I'm writing investor updates, and seeing an author who...wants to treat writing a book like it's a startup and a book advance like a raise...neat!
I was just reading about JSD the other day after reading about KL divergence...seems like a nifty measurement device for things like sim-to-real evaluations in robots (the reason I was going down this rabbit hole.)
I think the appeal over raw KL is that JSD behaves a bit nicer when the simulated and real distributions don't perfectly overlap...which is basically always true in the real world!