Book Image

Deep Learning from the Basics

By : Koki Saitoh
5 (1)
Book Image

Deep Learning from the Basics

5 (1)
By: Koki Saitoh

Overview of this book

Deep learning is rapidly becoming the most preferred way of solving data problems. This is thanks, in part, to its huge variety of mathematical algorithms and their ability to find patterns that are otherwise invisible to us. Deep Learning from the Basics begins with a fast-paced introduction to deep learning with Python, its definition, characteristics, and applications. You’ll learn how to use the Python interpreter and the script files in your applications, and utilize NumPy and Matplotlib in your deep learning models. As you progress through the book, you’ll discover backpropagation—an efficient way to calculate the gradients of weight parameters—and study multilayer perceptrons and their limitations, before, finally, implementing a three-layer neural network and calculating multidimensional arrays. By the end of the book, you’ll have the knowledge to apply the relevant technologies in deep learning.

Related Content you might be interested in

Current Title:

Small book image
Deep Learning from the Basics
Koki Saitoh
Mar, 2021 | 316 pages
Small book image
Practical Convolutional Neural Networks
Mohit Sewak | Md Rezaul Karim | Pradeep Pujari
Feb, 2018 | 218 pages
Small book image
Hands-On Computer Vision with TensorFlow 2
Benjamin Planche | Eliot Andres
May, 2019 | 372 pages
Small book image
Hands-On Convolutional Neural Networks with TensorFlow
Leonardo Araujo | Nimesh Patel | Giounona Tzanidou | Richard Burton | Iffat Zafar
Aug, 2018 | 272 pages
Table of Contents (11 chapters)
Preface
Preface
About the Book
1
Introduction
Introduction
Concept of this book
Acknowledgments
Free Chapter
2
1. Introduction to Python
1. Introduction to Python
What is Python?
Installing Python
Python Interpreter
Python Script Files
NumPy
Matplotlib
Summary
3
2. Perceptrons
2. Perceptrons
What Is a Perceptron?
Simple Logic Circuits
Implementing Perceptrons
Limitations of Perceptrons
Multilayer Perceptrons
From NAND to a Computer
Summary
4
3. Neural Networks
3. Neural Networks
From Perceptrons to Neural Networks
Activation Function
Calculating Multidimensional Arrays
Implementing a Three-Layer Neural Network
Designing the Output Layer
Identity Function and Softmax Function
Handwritten Digit Recognition
Summary
5
4. Neural Network Training
4. Neural Network Training
Learning from Data
Loss Function
Numerical Differentiation
Gradient
Implementing a Training Algorithm
Summary
6
5. Backpropagation
5. Backpropagation
Computational Graphs
Chain Rule
Backward Propagation
Implementing a Simple Layer
Implementing the Activation Function Layer
Implementing the Affine and Softmax Layers
Implementing Backpropagation
Summary
7
6. Training Techniques
6. Training Techniques
Updating Parameters
Initial Weight Values
Regularization
Validating Hyperparameters
Summary
8
7. Convolutional Neural Networks
7. Convolutional Neural Networks
Overall Architecture
The Convolution Layer
The Pooling Layer
Implementing the Convolution and Pooling Layers
Implementing a CNN
Visualizing a CNN
Typical CNNs
Summary
9
8. Deep Learning
8. Deep Learning
Making a Network Deeper
A Brief History of Deep Learning
Accelerating Deep Learning
Practical Uses of Deep Learning
The Future of Deep Learning
Summary
Appendix A
Appendix A
Computational Graph of the Softmax-with-Loss Layer
Summary

Validating Hyperparameters

A neural network uses many hyperparameters, as well as parameters such as weights and biases. The hyperparameters here include the number of neurons in each layer, batch size, the learning rate for updating parameters, and weight decay. Setting the hyperparameters to inappropriate values deteriorates the performance of the model. The values of these hyperparameters are very important, but determining them usually requires a lot of trial and error. This section describes how to search for hyperparameter values as efficiently as possible.

Validation Data

In the dataset we've used so far, the training data and test data are separate. The training data is used to train a network, while the test data is used to evaluate generalization performance. Thus, you can determine whether or not the network conforms too well only to the training data (that is, whether overfitting occurs) and how large the generalization performance is.

We will use various...

Previous Section
End of Section 5
Next Section