4 ms·
Visualizing Neural Networks
- bookmark1231 3y agoChris Olah has a similar but more in depth blog post like this: https://colah.github.io/posts/2014-03-NN-Manifolds-Topology/ https://colah.github.io/posts/2014-03-NN-Manifolds-Topology/
- photon_lines 3y agoNice - thank you! Very informative post as well -- I love topology and this looks like a god-send!
- RandomWorker 3y agoThis is basically week 3 of Andrew Ng neural networks and deep learning course on Coursera.
- puttycat 3y agoOff-topic: Substack's dark pattern attempts to maximize traction using intrusive pop-ups, text hiding, etc, is really awful. Such a text-hostile platform.
- naillo 3y agoQue substack becoming medium and 'stacksub' popping up with a new clean interface to take the users. Rinse and repeat and we never learn.
- Terr_ 3y ago*Cue, a correction spurred by my own intrusively embarrassing memory of a Scrabble game misstep. English also has "queue" (like a line of people or a certain hairstyle) but "que" generally means somebody is speaking Spanish. Anyway, I hope that some future generations manage to figure out systems or incentives that avoid or delay that kind of mission decay.
- svachalek 3y agoSadly you are observing the process of learning, it's just not the lesson you hoped they would take.
- qudat 3y agoI tried to capture this phenomenon in writing: https://bower.sh/bowers-law https://bower.sh/bowers-law
- rdedev 3y agoWould it be better if substack has two forms of the same article? Like have the full article available only for paid members while having another one for someone just checking it out. Leave it to the author to decide what to show in a particular view. Substack can restrict the size or types of content available in the free view of the article to reduce their server usage. I feel like this way the eventual enshittification could be delayed or even avoided. Of course this way might not be the best for getting YoY growth. And I'm not sure if any website provides such a provision
- genter 3y agoIt's particularly nasty to have it popup when you've already read a couple paragraphs and are starting to get invested in it. At least Washington Post and NYTimes are upfront with their barrier.
- jwilber 3y agoHere’s a visual, interactive explanation with d3: https://mlu-explain.github.io/neural-networks/ https://mlu-explain.github.io/neural-networks/
- samsartor 3y agoA while back I made a video along these lines, if anyone wants to see it animated: https://youtu.be/CfAL_cL3SGQ?si=tPnCzBjI0QVOpwCQ https://youtu.be/CfAL_cL3SGQ?si=tPnCzBjI0QVOpwCQ
- photon_lines 3y agoGreat video - thank you for this!!
- alok-g 3y agoThose interested in the topic may check my video here: https://m.youtube.com/watch?v=B3NuJt6ttOI https://m.youtube.com/watch?v=B3NuJt6ttOI It's much more visual and easy-going. I created it upon finding that several software developers were unable to understand in spite of many resources already available on the topic.