nice!! thats so sick the magnitude thing makes so much sense intuitively
heyy really cool video!! Can't believe your don't habe mor views.
I personaly feel like some of the visualt for the Network its self were a litle bit over the top. Howerver don't be discuragec.
The over all quality was great.
Keep up the good work.
Brilliant
watching this made me wonder what would happen if you would try to train a Neural Network to resemble a given image. Im sure it has been done already, but I might try to mess around with that later
Good one
i thought he was gonna train the neural network to rebuild images in its output range graph which would’ve been insane but i’m already seeing some problems with that idea
All programming is kind of alien to me. I've never been good at understanding it. This is magnitudes and magnitudes more difficult to understand. So I guess it all starts with guessing but how does a computer guess? RNG? What even is RNG related to starting from no seed?
RNG is pretty much always done by a seed, so the numbers can be consistent when needed, paired with something called a chaos function.
imagine just opening desmos and spamming random things until you get a function that is pretty much unpredictable (in reality its prob different)
you can also generate random numbers from pi and other such numbers :)
the RELU images look very similar to what happens when you combine two images with 100% opacity on 'hard mix' blending mode in image editing software!
Fascinating content.