CNN Note - Back Propagation Alogorithm

by lucainiaoge

---

Intro

This is a detailed derivation and summerization on back propagation (BP) algorithm in fully-connected neural networks and CNN:

References and helpful links

• Reference 1: CNN tutorial note of Jake Bouvrie
• Reference 2: A helpful CSDN blog in Chinese, author - xqp_dream

Prerequisites for this note

neural network concepts, CNN concepts, vector calculus

First Part: Review – BP (back propagation) in fully connected layers

Review of Feedforward and BP formula

Just a review, not a tutorial! How to derive the formula? Please go to learn neural network basics.
Give a link of tutorial from Andrew Ng

Detailed derivation of BP formula

Second Part: math in CNN – representation and BP

Math formalization of CNN feeding forward

BP for convolutional layers

The start of error terms and dealing with BP of the $L^{th}$ layer (which is a pooling layer)

Dealing with BP of the $(L-1)^{th}$ layer (which is a convolutional layer)

BP for CNN – a summarization

本文作者： lucainiaoge
本文链接： https://lucainiaoge.github.io.git/2020/04/03/CNNnote2-BP_algorithm/
版权声明： 本作品采用 Creative Commons authorship - noncommercial use - same way sharing 4.0 international license agreement 进行许可。转载请注明出处！

---

Welcome to my little world

---

赏

Thx!!