@@robo0428 So this is SEV #4... I clicked through a few of the recent videos on 3b1b, but ! can't find some with these SEVs. Would you know where the other three are? (I usually do watch until the end, but I never really look at the endscreen videos because they are usually ones I've watched before).
I wish I had the time to work on this coding problem. I might pick it up in a few weeks we'll see. I do want to say though, I absolutely **love** the idea of 2nd editions of your UA-cam content. I don't know about other creators but for your content specifically, it makes perfect sense. Love it. Also... more Matt Parker collabs por favor! 🤗
To find squares I'd do something similiar to newton's method and finding roots. Like you search for a rectangle with a certain aspect ratio, tryna tweak each points to get closer, and once you're close enough but didn't find a square, the math behind make you go very var away, like 2 randoms pair. I don't know about efficiency, but it should find something pretty close easily
Hi Grant! Thanks for all your videos!! This is a slightly inside baseball question but what does your process of translating the pure math into code look like? Are you sitting with a pen and paper for hours thinking about how to convert the theory into programming objects (and then deriving animations from them too!) + a live ipython REPL open to brainstorm ideas -- if so, how do you deal with testing and possible bugs in your implementation? I saw your Manim video with Ben Sparks where you go into the tool itself but your process of communicating mathematical thoughts to code really fascinates me as well.
On an actual side note, wonder what the space of quadrilateral looks like (or topologically is like) on smooth curves such that gradient descent does decent progress.
Gauss-Newton converges faster, but while likely less speedy to compute, damped least-squares would likely give closer answers than just gradient descent or gauss-newton alone. Personally, having each frame take a little bit longer to render is a decent tradeoff for having accurate visuals plus pleasing eagle-eyed commenters.
Wait, am I missing out on the other secret endscreen videos??? 🤯
yeah man. you gotta watch all the way to the end to catch the link.
@@robo0428 So this is SEV #4... I clicked through a few of the recent videos on 3b1b, but ! can't find some with these SEVs. Would you know where the other three are? (I usually do watch until the end, but I never really look at the endscreen videos because they are usually ones I've watched before).
"I like to write 2pi as tau" is so funny lol
I wish I had the time to work on this coding problem. I might pick it up in a few weeks we'll see. I do want to say though, I absolutely **love** the idea of 2nd editions of your UA-cam content. I don't know about other creators but for your content specifically, it makes perfect sense. Love it.
Also... more Matt Parker collabs por favor! 🤗
8:40
Grant: "it includes a _very_ special guest"
Me: imagines an animated phi creature
same
It's closer to a tau creature...
a terry tao creature?
@@GrantSandersonSteve mould
@@GrantSanderson So that means you've made transparent 2D animations of things?
Thanks so much for all your amazing work, Grant!
To find squares I'd do something similiar to newton's method and finding roots. Like you search for a rectangle with a certain aspect ratio, tryna tweak each points to get closer, and once you're close enough but didn't find a square, the math behind make you go very var away, like 2 randoms pair.
I don't know about efficiency, but it should find something pretty close easily
saw this notif and thought I somehow missed a vid on the main channel haha
Hi Grant! Thanks for all your videos!!
This is a slightly inside baseball question but what does your process of translating the pure math into code look like? Are you sitting with a pen and paper for hours thinking about how to convert the theory into programming objects (and then deriving animations from them too!) + a live ipython REPL open to brainstorm ideas -- if so, how do you deal with testing and possible bugs in your implementation? I saw your Manim video with Ben Sparks where you go into the tool itself but your process of communicating mathematical thoughts to code really fascinates me as well.
So I've just found out about the existence of the secret videos. I guess they are less of a secret now.
As a viewer of UA-cam on a smart TV, I had no idea these videos existed...
try casting from a phone maybe?
So nice. Thanks a lot! 😀
How do I get access to the other SEVs?
I’ve been let in on a secret I wasn’t supposed to find
I just realized is right eye is 3 BLUE 1 BROWN
Where are vlogs 1-3?
He just explained it at the start of the video... Which explains why you haven't found them, I guess :D
@@YakushiiI didn’t realize the other end screen videos are the vlogs
Unlisted doesn’t mean private. That means there’s some way to access rhen
@@Archimedes115 just need a playlist link to them
It's all gradient descent. Finding rectangles, finding the parametrization, proving the theorem.
Gradient descent all the way down.
On an actual side note, wonder what the space of quadrilateral looks like (or topologically is like) on smooth curves such that gradient descent does decent progress.
Gauss-Newton converges faster, but while likely less speedy to compute, damped least-squares would likely give closer answers than just gradient descent or gauss-newton alone. Personally, having each frame take a little bit longer to render is a decent tradeoff for having accurate visuals plus pleasing eagle-eyed commenters.
That transform amazes me. Make a loop short of it and see how many views it gets Grant!
Oooh It showed up on the top of my recommendations
Secret vlog gang
2:26 I have started doing this as well. It just feels much more right.
Thank you!
im am so dumb i can barely appreciate this video
I'm early too. Like before watching as always
UwU what's this?
... no I mean really, wtf is this?
Ey I'm early!
Huh, we're special
Hi early, i'm too.