|
Table Of Contents
1 Introduction
1.1 Features of This Package
2 Neural Network Theory—A Short Tutorial
2.1 Introduction to Neural Networks
2.1.1 Function Approximation
2.1.2 Time Series and Dynamic Systems
2.1.3 Classification and Clustering
2.2 Data Preprocessing
2.3 Linear Models
2.4 The Perceptron
2.5 Feedforward and Radial Basis Function Networks
2.5.1 Feedforward Neural Networks
2.5.2 Radial Basis Function Networks
2.5.3 Training Feedforward and Radial Basis Function Networks
2.6 Dynamic Neural Networks
2.7 Hopfield Network
2.8 Unsupervised and Vector Quantization (VQ) Networks
2.9 Further Reading
3 Getting Started and Basic Examples
3.1 Palettes and Loading the Package
3.1.1 Palettes
3.1.2 Loading the Package and Data
3.2 Package Conventions
3.2.1 Data Format
3.2.2 Function Names
3.2.3 Network Format
3.3 NetClassificationPlot
3.4 Basic Examples
3.4.1 Classification Problem Example
3.4.2 Function Approximation Example
4 The Perceptron
4.1 Perceptron Network Functions and Options
4.1.1 InitializePerceptron
4.1.2 PerceptronFit
4.1.3 NetInformation
4.1.4 NetPlot
4.2 Examples
4.2.1 Two Classes in Two Dimensions
4.2.2 Several Classes in Two Dimensions
4.2.3 Higher-Dimensional Classification
4.3 Further Reading
5 The Feedforward Neural Network
5.1 Feedforward Network Functions and Options
5.1.1 InitializeFeedForwardNet
5.1.2 NeuralFit
5.1.3 NetInformation
5.1.4 NetPlot
5.1.5 LinearizeNet and NeuronDelete
5.1.6 SetNeuralD, NeuralD, and NNModelInfo
5.2 Examples
5.2.1 Function Approximation in One Dimension
5.2.2 Function Approximation from One to Two Dimensions
5.2.3 Function Approximation in Two Dimensions
5.3 Classification with Feedforward Networks
5.4 Further Reading
6 The Radial Basis Function (RBF) Network
6.1 RBF Network Functions and Options
6.1.1 InitializeRBFNet
6.1.2 NeuralFit
6.1.3 NetInformation
6.1.4 NetPlot
6.1.5 LinearizeNet and NeuronDelete
6.1.6 SetNeuralD, NeuralD, and NNModelInfo
6.2 Examples
6.2.1 Function Approximation in One Dimension
6.2.2 Function Approximation from One to Two Dimensions
6.2.3 Function Approximation in Two Dimensions
6.3 Classification with RBF Networks
6.4 Further Reading
7 Training Feedforward and Radial Basis Function Networks
7.1 NeuralFit
7.2 Examples of Different Training Algorithms
7.3 Train with FindMinimum
7.4 Troubleshooting
7.5 Regularization and Stopped Search
7.5.1 Regularization
7.5.2 Stopped Search
7.5.3 Example
7.6 Separable Training
7.6.1 Small Example
7.6.2 Larger Example
7.7 Options Controlling Training Results Presentation
7.8 The Training Record
7.9 Writing Your Own Training Algorithms
7.10 Further Reading
8 Dynamic Neural Networks
8.1 Dynamic Network Functions and Options
8.1.1 Initializing and Training Dynamic Neural Networks
8.1.2 NetInformation
8.1.3 Predicting and Simulating
8.1.4 Linearizing a Nonlinear Model
8.1.5 NetPlot—Evaluate Model and Training
8.1.6 MakeRegressor
8.2 Examples
8.2.1 Identifying the Dynamic of a DC Motor
8.2.2 Identifying the Dynamics of a Hydraulic Actuator
8.2.3 Bias-Variance Tradeoff—Avoiding Overfitting
8.2.4 Fix Some Parameters—More Advanced Model Structures
8.3 Further Reading
9 Hopfield Networks
9.1 Hopfield Network Functions and Options
9.1.1 HopfieldFit
9.1.2 NetInformation
9.1.3 HopfieldEnergy
9.1.4 NetPlot
9.2 Examples
9.2.1 Discrete-Time Two-Dimensional Example
9.2.2 Discrete-Time Classification of Letters
9.2.3 Continuous-Time Two-Dimensional Example
9.2.4 Continuous-Time Classification of Letters
9.3 Further Reading
10 Unsupervised Networks
10.1 Unsupervised Network Functions and Options
10.1.1 InitializeUnsupervisedNet
10.1.2 UnsupervisedNetFit
10.1.3 NetInformation
10.1.4 UnsupervisedNetDistance, UnUsedNeurons, and NeuronDelete
10.1.5 NetPlot
10.2 Examples without SOM
10.2.1 Clustering in Two-Dimensional Space
10.2.2 Clustering in Three-Dimensional Space
10.2.3 Pitfalls with Skewed Data Density and Badly Scaled Data
10.3 Examples with SOM
10.3.1 Mapping from Two to One Dimensions
10.3.2 Mapping from Two Dimensions to a Ring
10.3.3 Adding a SOM to an Existing Unsupervised Network
10.3.4 Mapping from Two to Two Dimensions
10.3.5 Mapping from Three to One Dimensions
10.3.6 Mapping from Three to Two Dimensions
10.4 Change Step Length and Neighbor Influence
10.5 Further Reading
11 Vector Quantization
11.1 Vector Quantization Network Functions and Options
11.1.1 InitializeVQ
11.1.2 VQFit
11.1.3 NetInformation
11.1.4 VQDistance, VQPerformance, UnUsedNeurons, and NeuronDelete
11.1.5 NetPlot
11.2 Examples
11.2.1 VQ in Two-Dimensional Space
11.2.2 VQ in Three-Dimensional Space
11.2.3 Overlapping Classes
11.2.4 Skewed Data Densities and Badly Scaled Data
11.2.5 Too Few Codebook Vectors
11.3 Change Step Length
11.4 Further Reading
12 Application Examples
12.1 Classification of Paper Quality
12.1.1 VQ Network
12.1.2 RBF Network
12.1.3 Feedforward Network
12.2 Prediction of Currency Exchange Rate
13 Changing the Neural Network Structure
13.1 Change the Parameter Values of an Existing Network
13.1.1 Feedforward Network
13.1.2 RBF Network
13.1.3 Unsupervised Network
13.1.4 Vector Quantization Network
13.2 Fixed Parameters
13.3 Select Your Own Neuron Function
13.3.1 The Basis Function in an RBF Network
13.3.2 The Neuron Function in a Feedforward Network
13.4 Accessing the Values of the Neurons
13.4.1 The Neurons of a Feedforward Network
13.4.2 The Basis Functions of an RBF Network
|
