News for hackers. Stuff that matters.
Re:Visualizing Neural Networks by bookmark1231 on Friday August 25, 2023 @11:42AM (#37262983)
Chris Olah has a similar but more in depth blog post like this: https://colah.github.io/posts/2014-03-NN-Manifolds-Topology/
Re:Visualizing Neural Networks by photon_lines on Friday August 25, 2023 @02:01PM (#37264957)
Nice - thank you! Very informative post as well -- I love topology and this looks like a god-send!
Re:Visualizing Neural Networks by RandomWorker on Friday August 25, 2023 @11:43AM (#37263005)
This is basically week 3 of Andrew Ng neural networks and deep learning course on Coursera.
Re:Visualizing Neural Networks by puttycat on Friday August 25, 2023 @11:58AM (#37263192)
Off-topic: Substack's dark pattern attempts to maximize traction using intrusive pop-ups, text hiding, etc, is really awful. Such a text-hostile platform.
Re:Visualizing Neural Networks by naillo on Friday August 25, 2023 @01:04PM (#37264186)
Que substack becoming medium and 'stacksub' popping up with a new clean interface to take the users. Rinse and repeat and we never learn.
Re:Visualizing Neural Networks by Terr_ on Friday August 25, 2023 @02:54PM (#37265727)
*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.

Re:Visualizing Neural Networks by svachalek on Friday August 25, 2023 @04:27PM (#37266618)
Sadly you are observing the process of learning, it's just not the lesson you hoped they would take.
Re:Visualizing Neural Networks by qudat on Friday August 25, 2023 @08:11PM (#37268605)
I tried to capture this phenomenon in writing: https://bower.sh/bowers-law
Re:Visualizing Neural Networks by rdedev on Friday August 25, 2023 @04:40PM (#37266759)
Would 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

Re:Visualizing Neural Networks by genter on Friday August 25, 2023 @10:02PM (#37269266)
It'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.
Re:Visualizing Neural Networks by jwilber on Friday August 25, 2023 @12:39PM (#37263789)
Here’s a visual, interactive explanation with d3:

https://mlu-explain.github.io/neural-networks/

Re:Visualizing Neural Networks by samsartor on Friday August 25, 2023 @01:42PM (#37264677)
A while back I made a video along these lines, if anyone wants to see it animated: https://youtu.be/CfAL_cL3SGQ?si=tPnCzBjI0QVOpwCQ
Re:Visualizing Neural Networks by photon_lines on Friday August 25, 2023 @01:52PM (#37264816)
Great video - thank you for this!!
Re:Visualizing Neural Networks by alok-g on Friday August 25, 2023 @07:17PM (#37268225)
Those interested in the topic may check my video here:

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.