#How to implement this in Python only using Numpy

11 messages · Page 1 of 1 (latest)

safe willow
safe willow
vital mesa
#

What's your question?

safe willow
#

im trying to build a NN using only numpy and im getting confused with the back propagation part , even though all the theory asks of me to use the chain rule , all the python implementations seem to use different methods

#

ive already implemented the method in this article and it does work , but i have no idea of how we even got these expressions

vital mesa
safe willow
#

https://deeplizard.com/learn/video/Zr5viAZGndE#:~:text=individual training sample.-,To calculate the derivative of the loss with respect to,function over all training samples.&text=∂ C ∂ w 12,∂ w 12 ( L ) sorry this is what i am finding hard to implement in python , because the shapes of matrices are not matching when i do the dot product , even though i followed the chain rule as in

Backpropagation explained | Part 4 - Calculating the gradient

We're now on number 4 in our journey through understanding backpropagation. In our last video, we focused on how we can mathematically express certain facts about the training process. Now we're

vital mesa
#

Okay?

#

So what specifically?

safe willow
#

i tried direclty pluggin in this method into a python function , but the matrice shapes werent matching and also I have no idea how to find the derivative of the softmax function