template

What a stunning interactive visualization. I wonder what tool was used here? Edit: Here's their template https://github.com/distillpub/template/tree/master As I feared, they seem to be using D3.js "by hand", which is the only way to achieve such high quality but needs IMO an enormous amount of work and dedication. No wonder they are offline, that's hard to sustain. I hope a technology comes that makes it significantly easier to make such interactive charts. I know there are many out there, but there's always a "but".

I've previously read some in-depth articles on ML & DL that used Distill's publishing templates. Might be worth a look. https://github.com/distillpub/template

It's looks like it's based on Distill's template (distill.pub), which is available here: https://github.com/distillpub/template

It's under review for Distill (there's a mention of it on the page as well). It's using the Distill template - https://github.com/distillpub/template

You're right: we are taking the mean of the activations of a given channel `z` over all its `x,y` coordinates. (We could sum, but we use mean so that step sizes are comparable between channel and neuron objectives.) Thanks for the feedback that this notation is not super clear, we will consider rewriting those expressions. When we do feature visualization we do start from a random point/noise. For the diagram showing steepest descent directions, however, the gradient is evaluated on an input image from the dataset, shown as the leftmost image. There's no real step size either as we're showing the direction. You can think of the scale as arbitrary and chosen for appearance. Section numbers are on their way—and figure numbers also sound helpful! I've added a ticket. (https://github.com/distillpub/template/issues/63) For now you can alread

Thanks for the suggestion, added it to our issues... https://github.com/distillpub/template/issues/62

Looking at the finger instead of the moon: I like the HTML layout (responsive, inline images with captions, lateral notes). Any insights on how it's generated? Markdown, Rst, Latex -> HTML? I would love to produce my documentation in this way. Edit: I was too hurried. Everything is explained in https://distill.pub/guide/, the template is at https://github.com/distillpub/template

We were frustrated by the experience of reading machine learning papers on screens (particularly phones/tablets). There are lots of good tools for authoring HTML papers (Distill, Authorea, etc) but nothing that deals with the vast number of PDF papers that already exist. So, we built Arxiv Vanity: a site that renders Arxiv papers as web pages. It’s still pretty janky, but for the papers that do render correctly, the experience is so much better than reading a PDF. For example: https://www.arxiv-vanity.com/papers/1705.04085v3/ https://www.arxiv-vanity.com/papers/1708.00884/ https://www.arxiv-vanity.com/papers/1705.06031v2/ The source for the LaTeX to HTML renderer is on GitHub[0]. It’s built on Pandoc[1] and Distill.pub’s template[2]. [0] https://github.com/arxiv-vanity/engrafo [1] https://pandoc.org [2] https://github.com/distillpub/templ

They did actually! [0] The blog posts are also online on their GitHub site. [0] https://github.com/distillpub/template