The Backpropagation algorithm for a math student

A Deep Neural Network (DNN) is a composite function of vector-valued functions, and in order to train a DNN, it is necessary to calculate the gradient of the loss function with respect to all parameters. This calculation can be a non-trivial task because the loss function of a DNN is a composition o...

Full description

Bibliographic Details

Published in:	Proceedings of ... International Joint Conference on Neural Networks pp. 01 - 09
Main Authors:	Damadi, Saeed, Moharrer, Golnaz, Cham, Mostafa, Shen, Jinglai
Format:	Conference Proceeding
Language:	English
Published:	IEEE 18.06.2023
Subjects:	Complexity theory Computer architecture Convolutional neural networks Iterative algorithms Jacobian matrices Stochastic processes Transformers
ISSN:	2161-4407
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!