Considering that back propagation is an important algorithm, and derivation is still difficult, SO I plan to write a good derivation of the algorithm. It is not easy to write as succinctly and collociously as possible. Considering that many people will use Ng’s Machine Learning course [1], my symbols will follow the course’s symbols.
This course did not derive back propagation algorithm, but I think it is necessary to learn its derivation during self-study and consolidation, otherwise it can not be regarded as a true knowledge of the principle of BP. In writing this book, I deepened my impression by eschewing all references and doing the derivation entirely on my own, and now I intend to present the derivation in a rigorous and colloquial way.
The book has been released and can be found on the DezemingFamily website dezeming. Top /.
Some screenshots of the book:
Screenshot 2:
Screenshots:
