Notes for t-SNE paper

gist.github.com

Nice, but a few pictures would be really helpful, even for 2d -> 2d.

(I know how does t-SNE work, but I am looking for a simple material for others. Especially showing what happens when you adjust perplexity.)

You may find this js implementation pretty useful https://github.com/karpathy/tsnejs

I know this one. I was the second time I wanted to code something in ML in JS, but already saw Karpathy did it.

Haha. He seems to have a special place for js in his heart :)

I think it is more that he loves exposing ML, and is great at it.