Backpropagation: how it works

preview_player
Показать описание

Рекомендации по теме
Комментарии
Автор

Thanks for the explanation! It's the clearest explanation I have found so far.

alexanderkorsunsky
Автор

Really liked your short videos. Thanks!

fukeya
Автор

Amazing video... simple and yet not oversimplified. Thank you very much for uploading.

dimitrab
Автор

Undoubtedly all of your videos are excellent and very comprehensible. It would be very nice if you could add some information about the topics that your cover in each videos So we can find right video easily.
Thank you.

hosseinpourghaemi
Автор

Thank you for video, it's was very easy to understand!

tommyunreal
Автор

Good day Sir. Can you please clarify your interpretation of the sigmoid function, 4:00 - 4:20? I can't hear it quite well.
This vid helped me out too, thank you.

arahflorgallardo
Автор

what exactly does a node do with for example 4 inputs? does it add them together and divide by 4 then multiply by a number? what do the weights do? is it something you do to the input number or does it tell the node something?

minecraftermad
Автор

At 5:15 Victor says "f.k times u.jk is just one component of the sum that feeds into g.j, so I just need to differentiate".

Could anyone explain that comment about differentiating? Why does that help us?

MrAgreeandDisagree
Автор

is dE/dhi outside or inside of the summation?

RigatoniModular
Автор

how are weights of edges computed and how do we know them before hand ?

abdul
Автор

Thank you for the video.

I'm sorry, I do not follow.

If g = sigma(mu + sum (mu_k*h_k) then

dg/d(h_k) = sigma'(h_k)*mu_k, which means that dg = sigma'(h_k)*mu_k * dh_k

so dE/dg = (just inserting) dE/((sigma'(h)*mu_k * dh_k ) = 1/(sigma'(h_k)*mu_k * dh_k ) *dE/d(h_k), so NOT equal to

(sigma'(h)*mu_k * dh_k ) *dE/d(h_k)

??

NisseOhlsen
Автор

too bad i havent gotten to that state of mathematics in school yet

minecraftermad