Thank you so much for making this, I was already familiar with backprop butt here were too many doubts of the inner working but this video made it clear. Again thank you!
This is a superb video, EXCEPT that I think there's a mistake in the very first calculation. Shouldn't the sums be i1*w1 + i2*w2 and i1*w3 + i2*w4 ?? Please double check this; forgive me if I'm wrong.
correct me if I am wrong, but this is how I would explain it: y is not the actual image of a dog. It's a scalar, meaning "just" a "normal" number or value. it serves as an indicator. The image of a dog is nothing more but a matrix and this matrix contains certain numbers (or vectors). And the image of a dog can only be categorized as such if you also have images labled "not dog" to train the algorithm. in the end y and y' are nothing but abstract values to determine a difference between what the model has learned about the image of a dog and what an actual image of a dog looks like. if the difference between y and y' is relatively small, then this means that the algorithm is trained well for this case and might recognize images of a dog in the future.
Don’t click on random video kids, or else you’ll learn about neural networks and calculus. Very scary.
LOL
😂
Always a random small yt channel has the best learning videos
Something I find crazy is that googles networks identified my video as having two main chapters, and MARKED THEM ACCORDINGLY.
there is always one guy that will give you what you need in 15 min . after you searched the TH-cam for 5 Hours. ( and found no thing )
This is truly excellent video i finally understand something after going through numerous videos
Super great video. You are great at teaching. I hope you are a professor. Thank you, best explanation by far.
Thank you so much for making this, I was already familiar with backprop butt here were too many doubts of the inner working but this video made it clear. Again thank you!
Watched some videos on backpropagation and yours is the most clear. Thanks
....this is a fantastic video...u really explained this well...
The best out here
Thank you for such a great explanation!
This is a superb video, EXCEPT that I think there's a mistake in the very first calculation. Shouldn't the sums be i1*w1 + i2*w2 and i1*w3 + i2*w4 ?? Please double check this; forgive me if I'm wrong.
Me too
Is it a mistake?
It's more like
(net)h1 = i1*w1 + i2*w3
(net)h2 = i1*w2 + i2*w4
that is an art
it's amazing how simple you made it sound. can you do a version for networks with multiple hidden layers please ?
In 6:15, From the equation perspective, what is the value of y if there is a dog in the image?
correct me if I am wrong, but this is how I would explain it:
y is not the actual image of a dog. It's a scalar, meaning "just" a "normal" number or value. it serves as an indicator.
The image of a dog is nothing more but a matrix and this matrix contains certain numbers (or vectors). And the image of a dog can only be categorized as such if you also have images labled "not dog" to train the algorithm.
in the end y and y' are nothing but abstract values to determine a difference between what the model has learned about the image of a dog and what an actual image of a dog looks like. if the difference between y and y' is relatively small, then this means that the algorithm is trained well for this case and might recognize images of a dog in the future.
Nice!
@Orblitz , These slides are amazing. Can I have the link of this slide please. Want to print it out and keep for revision time to time.
Netflix level Content
Why can't you just do the chain rule?
fix your title: What is calculios, backpropagiation ?
Thanks lol, Im not the brightest at english