News for hackers. Stuff that matters.
Re:Visual explanations of core machine learning concepts by codedeadlock on Tuesday May 24, 2022 @06:07AM (#31490094)
This is neat and reminds me of R2D3 articles [1], how are these visualisations made? I know the hard way is hand-coding every aspect with D3, is there any simple framework that helps out in creating visual stories like these?

[1] - http://www.r2d3.us/visual-intro-to-machine-learning-part-1/

Re:Visual explanations of core machine learning concepts by jwilber on Tuesday May 24, 2022 @01:11PM (#31494805)
Each article's specific resources are listed at the bottom of the articles, but you're right - right now the articles are made using D3, IntersectionObserver API (for scrolling), and vanilla js.

Newer articles use svelte instead of vanilla js, for the following reasons: declarative markup makes constructing charts easier, component-scoped styles make collaboration easier, reactivity allows easy sharing of events between charts, and responsiveness is much, much easier (reactive width/height).

No simple framework exists yet per se, they pretty much all require getting down and dirty on the dataviz side of things. That said, there is some good progress on the interactive article side, the best of which I can think of is Matthew Conlen's Idyll: https://idyll-lang.org/docs

Re:Visual explanations of core machine learning concepts by lvl102 on Tuesday May 24, 2022 @07:18AM (#31490571)
Have you tried Vega? It sits on top of D3.