Add a batch normalization layer with its back propagation about cnn-from-scratch HOT 9 CLOSED

Backnorm? you mean batch_norm? or back_prop?
either one!! None of it depends on the type of dataset. Back_prop depends on Computation Graph and Batch_norm depends on direct input image.

from cnn-from-scratch.

Basically you want to add batch norm ?

from cnn-from-scratch.

Yes

from cnn-from-scratch.

I can help in adding the batch_norm layer but not sure about its back_prop part..
Can someone help?

from cnn-from-scratch.

cs231n Backpropagation
A little implementation part of the same

Above links would help you to implement backpropagation with any computation graph(In this case it is Batch_Norm)

from cnn-from-scratch.

The assignments of CS231N have the batch_norm setup for CIFAR 10

from cnn-from-scratch.

Hi, I have implemented the forward and the backwards pass of batch Norm layer in python using numpy but it's in function format, Will it help?
Will try to incorporate it into you code style
Is this what you want?

from cnn-from-scratch.

Yes, you can integrate that into the convnet() function.

from cnn-from-scratch.

Will try my best

from cnn-from-scratch.