
Deep Neural Networks in a Mathematical Framework
Description
Alles über E-Books | Antworten auf Fragen rund um E-Books, Kopierschutz und Dateiformate finden Sie in unserem Info- & Hilfebereich.
This SpringerBrief describes how to build a rigorous end-to-end mathematical framework for deep neural networks. The authors provide tools to represent and describe neural networks, casting previous results in the field in a more natural light. In particular, the authors derive gradient descent algorithms in a unified way for several neural network structures, including multilayer perceptrons, convolutional neural networks, deep autoencoders and recurrent neural networks. Furthermore, the authors developed framework is both more concise and mathematically intuitive than previous representations of neural networks.
This SpringerBrief is one step towards unlocking the black box of Deep Learning. The authors believe that this framework will help catalyze further discoveries regarding the mathematical properties of neural networks.This SpringerBrief is accessible not only to researchers, professionals and students working and studying in the field of deep learning, but alsoto those outside of the neutral network community.
More details
Other editions
Additional editions

Content
2 - Contents [Seite 10]
3 - Acronyms [Seite 12]
4 - 1 Introduction and Motivation [Seite 13]
4.1 - 1.1 Introduction to Neural Networks [Seite 14]
4.1.1 - 1.1.1 Brief History [Seite 14]
4.1.2 - 1.1.2 Tasks Where Neural Networks Succeed [Seite 15]
4.2 - 1.2 Theoretical Contributions to Neural Networks [Seite 16]
4.2.1 - 1.2.1 Universal Approximation Properties [Seite 16]
4.2.2 - 1.2.2 Vanishing and Exploding Gradients [Seite 17]
4.2.3 - 1.2.3 Wasserstein GAN [Seite 18]
4.3 - 1.3 Mathematical Representations [Seite 19]
4.4 - 1.4 Book Layout [Seite 19]
4.5 - References [Seite 20]
5 - 2 Mathematical Preliminaries [Seite 23]
5.1 - 2.1 Linear Maps, Bilinear Maps, and Adjoints [Seite 24]
5.2 - 2.2 Derivatives [Seite 25]
5.2.1 - 2.2.1 First Derivatives [Seite 25]
5.2.2 - 2.2.2 Second Derivatives [Seite 26]
5.3 - 2.3 Parameter-Dependent Maps [Seite 27]
5.3.1 - 2.3.1 First Derivatives [Seite 28]
5.3.2 - 2.3.2 Higher-Order Derivatives [Seite 28]
5.4 - 2.4 Elementwise Functions [Seite 29]
5.4.1 - 2.4.1 Hadamard Product [Seite 30]
5.4.2 - 2.4.2 Derivatives of Elementwise Functions [Seite 31]
5.4.3 - 2.4.3 The Softmax and Elementwise Log Functions [Seite 32]
5.5 - 2.5 Conclusion [Seite 34]
5.6 - References [Seite 34]
6 - 3 Generic Representation of Neural Networks [Seite 35]
6.1 - 3.1 Neural Network Formulation [Seite 36]
6.2 - 3.2 Loss Functions and Gradient Descent [Seite 37]
6.2.1 - 3.2.1 Regression [Seite 37]
6.2.2 - 3.2.2 Classification [Seite 38]
6.2.3 - 3.2.3 Backpropagation [Seite 39]
6.2.4 - 3.2.4 Gradient Descent Step Algorithm [Seite 40]
6.3 - 3.3 Higher-Order Loss Function [Seite 41]
6.3.1 - 3.3.1 Gradient Descent Step Algorithm [Seite 44]
6.4 - 3.4 Conclusion [Seite 45]
6.5 - References [Seite 46]
7 - 4 Specific Network Descriptions [Seite 47]
7.1 - 4.1 Multilayer Perceptron [Seite 48]
7.1.1 - 4.1.1 Formulation [Seite 48]
7.1.2 - 4.1.2 Single-Layer Derivatives [Seite 49]
7.1.3 - 4.1.3 Loss Functions and Gradient Descent [Seite 50]
7.2 - 4.2 Convolutional Neural Networks [Seite 52]
7.2.1 - 4.2.1 Single Layer Formulation [Seite 52]
7.2.1.1 - Cropping and Embedding Operators [Seite 53]
7.2.1.2 - Convolution Operator [Seite 55]
7.2.1.3 - Max-Pooling Operator [Seite 58]
7.2.1.4 - The Layerwise Function [Seite 61]
7.2.2 - 4.2.2 Multiple Layers [Seite 62]
7.2.3 - 4.2.3 Single-Layer Derivatives [Seite 62]
7.2.4 - 4.2.4 Gradient Descent Step Algorithm [Seite 63]
7.3 - 4.3 Deep Auto-Encoder [Seite 64]
7.3.1 - 4.3.1 Weight Sharing [Seite 64]
7.3.2 - 4.3.2 Single-Layer Formulation [Seite 65]
7.3.3 - 4.3.3 Single-Layer Derivatives [Seite 66]
7.3.4 - 4.3.4 Loss Functions and Gradient Descent [Seite 67]
7.4 - 4.4 Conclusion [Seite 69]
7.5 - References [Seite 70]
8 - 5 Recurrent Neural Networks [Seite 71]
8.1 - 5.1 Generic RNN Formulation [Seite 71]
8.1.1 - 5.1.1 Sequence Data [Seite 72]
8.1.2 - 5.1.2 Hidden States, Parameters, and Forward Propagation [Seite 72]
8.1.3 - 5.1.3 Prediction and Loss Functions [Seite 74]
8.1.4 - 5.1.4 Loss Function Gradients [Seite 74]
8.1.4.1 - Prediction Parameters [Seite 75]
8.1.4.2 - Real-Time Recurrent Learning [Seite 76]
8.1.4.3 - Backpropagation Through Time [Seite 77]
8.2 - 5.2 Vanilla RNNs [Seite 82]
8.2.1 - 5.2.1 Formulation [Seite 82]
8.2.2 - 5.2.2 Single-Layer Derivatives [Seite 83]
8.2.3 - 5.2.3 Backpropagation Through Time [Seite 84]
8.2.4 - 5.2.4 Real-Time Recurrent Learning [Seite 86]
8.2.4.1 - Evolution Equation [Seite 86]
8.2.4.2 - Loss Function Derivatives [Seite 87]
8.2.4.3 - Gradient Descent Step Algorithm [Seite 88]
8.3 - 5.3 RNN Variants [Seite 88]
8.3.1 - 5.3.1 Gated RNNs [Seite 89]
8.3.2 - 5.3.2 Bidirectional RNNs [Seite 90]
8.3.3 - 5.3.3 Deep RNNs [Seite 90]
8.4 - 5.4 Conclusion [Seite 90]
8.5 - References [Seite 91]
9 - 6 Conclusion and Future Work [Seite 92]
9.1 - References [Seite 93]
10 - Glossary [Seite 94]
System requirements
File format: PDF
Copy protection: Watermark-DRM (Digital Rights Management)
System requirements:
- Computer (Windows; MacOS X; Linux): Use the free software Adobe Reader, Adobe Digital Editions, or any other PDF viewer of your choice (see eBook Help).
- Tablet/Smartphone (Android; iOS): Install the free app Adobe Digital Editions or another reading app for eBooks, e.g., PocketBook (see eBook Help).
- E-reader: Bookeen, Kobo, Pocketbook, Sony, Tolino and many more (only limited: Kindle).
The file format PDF always displays a book page identically on any hardware. This makes PDF suitable for complex layouts such as those used in textbooks and reference books (images, tables, columns, footnotes). Unfortunately, on the small screens of e-readers or smartphones, PDFs are rather annoying, requiring too much scrolling.
This eBook uses Watermark-DRM, a „soft” copy protection. This means that there are no technical restrictions to prevent illegal distribution. However, there is a personalised watermark embedded in the eBook that can be used to identify the purchaser of the eBook in the event of misuse and to provide evidence for legal purposes.
For more information, see our eBook Help page.