Download presentation
Presentation is loading. Please wait.
2
Forward and Backward Max Pooling
3
Upsampling in Pooling
4
Convolution Operation (Forward Pass)
5
Input Size : 3x3, Filter Size : 2x2, Output Size : 2x2
Output Equations
6
Derivative Computation (Backward Pass)
Final derivatives after performing back propagation Note: no deconvolution, we are trying to compute an update of weight
7
For a full derivation, go through this site
Only Numpy: Understanding Back Propagation for Transpose Convolution in Multi Layer CNN with Example and Interactive Code. back-propagation-for-transpose-convolution-in-multi-layer- cnn-with-c0a07d191981
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.