Practical MATLAB Deep Learning – A Project-Based Approach

Practical MATLAB Deep Learning – A Project-Based Approach
اسم المؤلف
Michael Paluszek, Stephanie Thomas
التاريخ
1 نوفمبر 2021
المشاهدات
التقييم
(لا توجد تقييمات)
Loading...

Practical MATLAB Deep Learning – A Project-Based Approach
Michael Paluszek, Stephanie Thomas
Contents
About the Authors XI
About the Technical Reviewer XIII
Acknowledgements XV
1 What Is Deep Learning? 1
1.1 Deep Learning . 1
1.2 History of Deep Learning 2
1.3 Neural Nets . 4
1.3.1 Daylight Detector . 8
1.3.2 XOR Neural Net . 9
1.4 Deep Learning and Data 16
1.5 Types of Deep Learning 18
1.5.1 Multilayer Neural Network 18
1.5.2 Convolutional Neural Networks (CNN) . 18
1.5.3 Recurrent Neural Network (RNN) 18
1.5.4 Long Short-Term Memory Networks (LSTMs) . 19
1.5.5 Recursive Neural Network 19
1.5.6 Temporal Convolutional Machines (TCMs) . 19
1.5.7 Stacked Autoencoders 19
1.5.8 Extreme Learning Machine (ELM) 19
1.5.9 Recursive Deep Learning . 19
1.5.10 Generative Deep Learning 20
1.6 Applications of Deep Learning . 20
1.7 Organization of the Book 22
2 MATLAB Machine Learning Toolboxes 25
2.1 Commercial MATLAB Software . 25
2.1.1 MathWorks Products . 25
2.2 MATLAB Open Source 27
2.2.1 Deep Learn Toolbox . 28
2.2.2 Deep Neural Network . 28
IIICONTENTS
2.2.3 MatConvNet . 28
2.2.4 Pattern Recognition and Machine Learning Toolbox (PRMLT) . 28
2.3 XOR Example . 28
2.4 Training . 37
2.5 Zermelo’s Problem . 38
3 Finding Circles with Deep Learning 43
3.1 Introduction . 43
3.2 Structure 43
3.2.1 imageInputLayer . 44
3.2.2 convolution2dLayer . 44
3.2.3 batchNormalizationLayer . 46
3.2.4 reluLayer . 46
3.2.5 maxPooling2dLayer . 47
3.2.6 fullyConnectedLayer . 48
3.2.7 softmaxLayer . 49
3.2.8 classificationLayer 49
3.2.9 Structuring the Layers 50
3.3 Generating Data: Ellipses and Circles . 51
3.3.1 Problem 51
3.3.2 Solution 51
3.3.3 How It Works . 51
3.4 Training and Testing 55
3.4.1 Problem 55
3.4.2 Solution 56
3.4.3 How It Works . 56
4 Classifying Movies 65
4.1 Introduction . 65
4.2 Generating a Movie Database . 65
4.2.1 Problem 65
4.2.2 Solution 65
4.2.3 How It Works . 65
4.3 Generating a Movie Watcher Database . 68
4.3.1 Problem 68
4.3.2 Solution 68
4.3.3 How It Works . 68
4.4 Training and Testing 70
4.4.1 Problem 70
4.4.2 Solution 70
4.4.3 How It Works . 71
IVCONTENTS
5 Algorithmic Deep Learning 77
5.1 Building a Detection Filter . 81
5.1.1 Problem 81
5.1.2 Solution 81
5.1.3 How It Works . 82
5.2 Simulating Fault Detection . 84
5.2.1 Problem 84
5.2.2 Solution 84
5.2.3 How It Works . 84
5.3 Testing and Training 87
5.3.1 Problem 87
5.3.2 Solution 87
5.3.3 How It Works . 88
6 Tokamak Disruption Detection 91
6.1 Introduction . 91
6.2 Numerical Model 93
6.2.1 Dynamics . 93
6.2.2 Sensors 96
6.2.3 Disturbances . 96
6.2.4 Controller . 98
6.3 Dynamical Model 100
6.3.1 Problem 100
6.3.2 Solution 100
6.3.3 How It Works . 100
6.4 Simulate the Plasma 102
6.4.1 Problem 102
6.4.2 Solution 102
6.4.3 How It Works . 103
6.5 Control the Plasma . 104
6.5.1 Problem 104
6.5.2 Solution 106
6.5.3 How It Works . 106
6.6 Training and Testing 107
6.6.1 Problem 107
6.6.2 Solution 107
6.6.3 How It Works . 108
7 Classifying a Pirouette 115
7.1 Introduction . 115
7.1.1 Inertial Measurement Unit 117
7.1.2 Physics 118
VCONTENTS
7.2 Data Acquisition 120
7.2.1 Problem 120
7.2.2 Solution 120
7.2.3 How It Works . 121
7.3 Orientation . 126
7.3.1 Problem 126
7.3.2 Solution 126
7.3.3 How It Works . 126
7.4 Dancer Simulation . 128
7.4.1 Problem 128
7.4.2 Solution 128
7.4.3 How It Works . 128
7.5 Real-Time Plotting . 132
7.5.1 Problem 132
7.5.2 Solution 132
7.5.3 How It Works . 132
7.6 Quaternion Display . 134
7.6.1 Problem 134
7.6.2 Solution 135
7.6.3 How It Works . 135
7.7 Data Acquisition GUI 138
7.7.1 Problem 138
7.7.2 Solution 138
7.7.3 How It Works . 138
7.8 Making the IMU Belt 146
7.8.1 Problem 146
7.8.2 Solution 146
7.8.3 How It Works . 146
7.9 Testing the System . 147
7.9.1 Problem 147
7.9.2 Solution 147
7.9.3 How It Works . 147
7.10 Classifying the Pirouette 149
7.10.1 Problem 149
7.10.2 Solution 149
7.10.3 How It Works . 150
7.11 Hardware Sources 154
8 Completing Sentences 155
8.1 Introduction . 155
8.1.1 Sentence Completion . 155
8.1.2 Grammar . 156
VICONTENTS
8.1.3 Sentence Completion by Pattern Recognition 157
8.1.4 Sentence Generation . 157
8.2 Generating a Database of Sentences 157
8.2.1 Problem 157
8.2.2 Solution 157
8.2.3 How It Works . 157
8.3 Creating a Numeric Dictionary . 159
8.3.1 Problem 159
8.3.2 Solution 159
8.3.3 How It Works . 159
8.4 Map Sentences to Numbers . 160
8.4.1 Problem 160
8.4.2 Solution 160
8.4.3 How It Works . 160
8.5 Converting the Sentences . 161
8.5.1 Problem 161
8.5.2 Solution 161
8.5.3 How It Works . 162
8.6 Training and Testing 163
8.6.1 Problem 163
8.6.2 Solution 164
8.6.3 How It Works . 164
9 Terrain-Based Navigation 169
9.1 Introduction . 169
9.2 Modeling Our Aircraft . 169
9.2.1 Problem 169
9.2.2 Solution 169
9.2.3 How It Works . 169
9.3 Generating a Terrain Model 177
9.3.1 Problem 177
9.3.2 Solution 177
9.3.3 How It Works . 177
9.4 Close Up Terrain 182
9.4.1 Problem 182
9.4.2 Solution 182
9.4.3 How It Works . 182
9.5 Building the Camera Model 183
9.5.1 Problem 183
9.5.2 Solution 183
9.5.3 How It Works . 184
9.6 Plot Trajectory over an Image . 187
VIICONTENTS
9.6.1 Problem 187
9.6.2 Solution 187
9.6.3 How It Works . 187
9.7 Creating the Test Images 190
9.7.1 Problem 190
9.7.2 Solution 190
9.7.3 How It Works . 190
9.8 Training and Testing 193
9.8.1 Problem 193
9.8.2 Solution 193
9.8.3 How It Works . 193
9.9 Simulation 197
9.9.1 Problem 197
9.9.2 Solution 197
9.9.3 How It Works . 197
10 Stock Prediction 203
10.1 Introduction . 203
10.2 Generating a Stock Market . 203
10.2.1 Problem 203
10.2.2 Solution 203
10.2.3 How It Works . 203
10.3 Create a Stock Market . 207
10.3.1 Problem 207
10.3.2 Solution 208
10.3.3 How It Works . 208
10.4 Training and Testing 210
10.4.1 Problem 210
10.4.2 Solution 210
10.4.3 How It Works . 210
11 Image Classification 219
11.1 Introduction . 219
11.2 Using a Pretrained Network 219
11.2.1 Problem 219
11.2.2 Solution 219
11.2.3 How It Works . 219
12 Orbit Determination 227
12.1 Introduction . 227
12.2 Generating the Orbits 227
12.2.1 Problem 227
VIIICONTENTS
12.2.2 Solution 227
12.2.3 How It Works . 227
12.3 Training and Testing 234
12.3.1 Problem 234
12.3.2 Solution 234
12.3.3 How It Works . 235
12.4 Implementing an LSTM 239
12.4.1 Problem 239
12.4.2 Solution 239
12.4.3 How It Works . 239
12.5 Conic Sections . 243
Bibliography 247
Index 24
Index
A
Aircraft model, 169, 170
Algorithmic Deep Learning Neural
Network (ADLNN), 77, 78
air turbine, 77, 79
AirTurbineSim.m, 79, 80
algorithmic filter/estimator,
80
pressure regulator input, 81
B
Bidirectional long short-term memory
(biLSTM), 107, 152, 213,
241
C
Camera model, 183
Classify function, 57
Commercial software, 25–27
Convolutional network
layer types
batchNormalizationLayer, 46
classificationLayer, 49
convolution2dLayer, 44–46
fullyConnectedLayer, 48
imageInputLayer, 44
maxPooling2dLayer, 47
reluLayer, 46–48
softmaxLayer, 49
one-set, window, 63
structuring, 50
Convolutional neural networks (CNN),
18, 28, 193
Convolution process, 45
Cross-entropy loss, 49
D
Dancer simulation
RHSDancer.m, 128
data structure, 128
double pirouette, simulation of, 131
linear acceleration, 130
parameters, 129
Data acquisition CUI, 138–146
Data acquisition system, 147–149
Daylight detector, 8–9
Deep learning system
applications, 20–21
camera model, building, 183
complete sentences, 163
data, 16–18
defined, 1
detection filter, 22
history, 2–3
network, 24
orientation, 126–127
types, 18–20
Deep Learn Toolbox, 28
Deep Neural Network, 28
Detection filter
air turbine, failures, 81
DetectionFilter.m, 82, 83
reset action, 84
specific gain matrix, 82
time constant, 82
Diamagnetic energy, 92
E
Edge localized mode (ELM), 93, 96, 97
Ellipses and circles
generate images,
Ellipses and circles (cont.)
train and test, 55–62
ELM, see Extreme learning machine
(ELM)
Euler’s equation, 118
Exclusive-or (XOR), 2, 9
activation function, 11
DLXOR.m script, 28–29
feedforwardnet, 37
Gaussian noise, 37
GUI, 29–30
hidden layers, 35, 36
mean output error, 15, 16
network training
histogram, 33
performance, 31
state, 32
neural net, 35
regression, 34
tansig, 35
truth table and solution networks, 10
weights, expand, 12
XORDemo, 11, 14
XOR.m, 10–11
XORTraining.m, 12–13
Extreme learning machine (ELM), 19
F
Fault detection simulation
detection filter, 86
DetectionFilterSim, 84, 85
failed tachometer, 87
regulator, fail, 85, 86
fminsearch, 173
fullyConnectedLayer, 48, 213
G
Generative Deep Learning, 20, 157
H
Handwriting analysis, 20
Hessian matrix, 37, 38
I
Image classification, 217
Image recognition, 20
IMU belt, 146–147
Inertial Measurement Unit (IMU),
117–118
International Tokamak Experimental
Reactor (ITER), 91
Joint European Torus (JET), 95
L
Levenberg Marquardt training algorithm, 37
Long short-term memory (LSTM) network,
19, 210, 239
lstmLayer (numHiddenUnits), 213
Lumped parameter model, 94
M
Machine learning, types, 2
Machine translation, 3, 20
Magnetohydrodynamic (MHD), 92
MatConvNet, 28
MathWorks products
Computer Vision System Toolbox, 27
Deep Learning toolbox, 26
Image Acquisition Toolbox, 27
Instrument Control Toolbox, 26
Parallel Computing Toolbox, 27
Statistics and Machine Learning
Toolbox, 26
Text Analytics Toolbox, 27
visualization tools, 25
Movie database
characteristics, 66
function demo, 67
generate, 65–68
viewer database, 71
Movie watchers
generate, 68–70
training window, 74, 76
Multilayer network, 1–3
250
J, KINDEX
N
Neural nets
neuron, 4
activation functions, 5, 6
LinearNeuron.m., 6, 7
threshold function, 7
two input, 4
Neural network research, 1
O
Open source tools, 27–28
Orbit determination
conic sections, 243–245
generation
Elliptical orbit, 229
Keplerian elements, 230–233
orbital motion, 229
test orbit, 234
theta, 228
two conics, a circle and ellipse, 227
LSTM, implementation, 239–242
test results, 242
training window, 242
validation data, 240
xTrain, 239–240
training and testing, 235–238
P
Patternnet network, 73
input/output, 75
training window, 76
Pattern Recognition and Machine Learning
Toolbox (PRMLT), 28
Pirouette, 115
baseball pitcher’s pitch, 116
center of mass, dancer, 119
classification, 149–150
bilstmLayer, 152
DancerNN.m, 150–151
neural net training, 154
testing neural network, 153
data acquisition
BluetoothTest.m, 124–125
communication state status, 122
instrumental control toolbox, 121
Mac dongle, 121
MATLAB Bluetooth function, 120
replying data, 122
IMU, 117–118
instrument control toolbox, 115
physics, 118–119
sources of hardware, 154
Q
Quadratic error, 11
Quaternion display
Ballerina.obj file, 135
dancer orientation, 138
QuaternionVisualization.m, 136
real time plots, 135
Quaternion operations, 126–127
R
rand, 16
randi, 57, 88
Real-time plotting, 131–134
Recurrent Neural Network (RNN), 18–19
Recursive Deep Learning, 19
regressionLayer, 213
Replaced recursive neural nets (RNNs),
19, 210
reshape, 17
Root-mean-square error (RMSE), 214, 241
S
sequenceInputLayer (inputSize), 213
Single-layer networks, 1, 2
Speech recognition, 20
Stacked autoencoders, 19
Stock prediction algorithm
generation
function PlotStock.m plots, 205
Geometric Brownian Motion, 203
high volatility, 207
multiple stocks, creation, 204
US stocks, 204
251INDEX
Stock prediction algorithm (cont.)
zero volatility, 206
stock market, creation, 208, 209
training and testing
bilstm layer, 216
LSTM layer, 210, 216, 217
neural net, layers, 213
predictAndUpdateState, 214, 215
RMSE, 214
RNNs, 210
stock price, 211, 212
training window, 214
Support Vector Machines (SVM), 3
Targeting, 20
Temporal convolutional machines
(TCMs), 19
TensorFlow, 3
Terrain-Based navigation
aircraft model
dynamical model, 172
fminsearch, 173
Gulfstream, 174
lift, drag, and gravity, 171, 172
North-East-Up coordinates,
velocity, 169, 170
numerical integration, 175
output, 176
trajectory, 177
camera model, building
Pinhole camera, 184, 185
source image and view, 186, 187
close up terrain, 182–183
generating terrain model, 177–181
Plot Trajectory, over image,
187–189
simulation
camera view and trajectory,
199
subplot, 197–198
terrain segments and aircraft path,
200, 201
test image, creation, 190–192
training and testing, 193–196
Testing and training
DetectionFilterNN.m, 88–89
faults, characterize, 87
feedforwardnet, 88
GUI, 89, 90
XOR problem, 87
Tokamaks disruptions
dynamical model, 99–102
factors, 91–93
numerical model
controller, 98–99
disturbances, 96–97
dynamics, 93–95
sensors, 96
plasma
control, 104, 106–107
simulation, 102–105, 108
train and test, 107–113
trainNetwork function, 56
Z
Zermelo’s problem
control angle, 40
cost, 41
costate equations, 40
defined, 38
Hamiltonian, 39
local and global minimums, 39
solutions, 41

كلمة سر فك الضغط : books-world.net
The Unzip Password : books-world.net

تحميل

يجب عليك التسجيل في الموقع لكي تتمكن من التحميل
تسجيل | تسجيل الدخول

التعليقات

اترك تعليقاً