[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

Search:


View post   

>> No.9412701 [View]
File: 22 KB, 500x388, 1436119899570.jpg [View same] [iqdb] [saucenao] [google]
9412701

>>9412603
Because the gasses in our atmosphere are blue in color
Its why things like mountains and shit have a blue tinge over them when they're fucking far away, or why the moon looks blue during the day

Getting real fucking sick of people over-complicating the answer, getting into color wavelength and light refraction shit
You may as well ask why anything is any color, if you're going there

>> No.9147706 [View]
File: 22 KB, 500x388, 1481421098628.jpg [View same] [iqdb] [saucenao] [google]
9147706

>>9136603
I get that not everybody has the opportunity to take a rigorous ML course but that course fucking sucks. Apparently knowing the chain rule makes you "an expert in calculus" according to Ng

>It turns out its faster to do it this way than using loops

Ng justifies nothing, just waves his hands and says "it turns out" that vectorization results in speedup and backprop just works.

Like the vectorization thing is so simple, the matrix operations are being computed in machine code, while things like for loops are being done interpreted by Python,

Here's a VERY good explanation of what NNs do mathematically:

http://colah.github.io/posts/2014-03-NN-Manifolds-Topology/

i realize im autistic af but ffs you even forgot the nonlinear transformation

Navigation
View posts[+24][+48][+96]