Backpropagation: How Neural Networks Learn

preview_player
Показать описание
A brief intro to the algorithm that powers virtually all neural network training today.

Timestamps
--------------------
Introduction 00:00
Neural network overview 00:31
Gradient descent 03:04
The backpropagation algorithm 05:05
Рекомендации по теме
Комментарии
Автор

Nicely done. Love the coloured up maths symbols. Realy makes it look more cuddly and easy.

markr
Автор

7:50 'say we have a neuron i connected to every neuron k in the following layer': proceeds to calculate dL/dyi ??? How does the output yi depend on the next weights ???
A few minutes before with the first example, you calculated dL/dyj ?

apianoadventure
Автор

what is xj hat ? why "xj hat = yj + wij"?

saminsolouki
Автор

You need to be more monotone, to help viewers to feel more excited about backprop.

pensiveintrovert