- Codeless Deep Learning with KNIME
- Why subscribe?
- Contributors
- About the authors
- About the reviewers
- Packt is searching for authors like you
- Preface
- Section 1: Feedforward Neural Networks and KNIME Deep Learning Extension
- Chapter 1: Introduction to Deep Learning with KNIME Analytics Platform
- The Importance of Deep Learning
- Exploring KNIME Software
- KNIME Analytics Platform
- KNIME Server for the Enterprise
- Exploring KNIME Analytics Platform
- Useful Links and Materials
- Build and Execute Your First Workflow
- Installing KNIME Deep Learning – Keras Integration
- Installing the Keras and TensorFlow Nodes
- Setting up the Python Environment
- Goal and Structure of this Book
- Summary
- Chapter 2: Data Access and Preprocessing with KNIME Analytics Platform
- Chapter 3: Getting Started with Neural Networks
- Neural Networks and Deep Learning – Basic Concepts
- Artificial Neuron and Artificial Neural Networks
- Signal Propagation within a Feedforward Neural Network
- Understanding the Need for Hidden Layers
- Training a Multilayer Perceptron
- Designing your Network
- Commonly Used Activation Functions
- Regularization Techniques to Avoid Overfitting
- Other Commonly used Layers
- Training a Neural Network
- Loss Functions
- Parameters and Optimization of the Training Algorithm
- Summary
- Questions and Exercises
- Chapter 4: Building and Training a Feedforward Neural Network
- Preparing the Data
- Datasets and Classification Examples
- Encoding of Nominal Features
- Normalization
- Other Helpful Preprocessing Nodes
- Data Preparation on the Iris Dataset
- Data Preparation on the Adult Dataset
- Building a Feedforward Neural Architecture
- The Keras Input Layer Node
- The Keras Dense Layer Node
- Building a Neural Network for Iris Flower Classification
- Building a Neural Network for Income Prediction
- Training the Network
- Selecting the Loss Function
- Defining the Input and Output Data
- Setting the Training Parameters
- Tracking the Training Progress
- Training Settings for Iris Flower Classification
- Training Settings for Income Prediction
- Testing and Applying the Network
- Executing the Network
- Extracting the Predictions and Evaluating the Network Performance
- Testing the Network Trained to Classify Iris Flowers
- Testing the Network Trained for Income Prediction
- Summary
- Questions and Exercises
- Section 2: Deep Learning Networks
- Chapter 5: Autoencoder for Fraud Detection
- Introducing Autoencoders
- Architecture of the Autoencoder
- Reducing the Input Dimensionality with an Autoencoder
- Detecting Anomalies Using an Autoencoder
- Why is Detecting Fraud so Hard?
- Building and Training the Autoencoder
- Data Access and Data Preparation
- Building the Autoencoder
- Training and Testing the Autoencoder
- Detecting Fraudulent Transactions
- Optimizing the Autoencoder Strategy
- Optimizing Threshold
- Threshold is defined on a separate subset of data, called the optimization set. There are two options here:
- Deploying the Fraud Detector
- Reading Network, New Transactions, and Normalization Parameters
- Applying the Fraud Detector
- Taking Actions
- Summary
- Questions and Exercises
- Chapter 6: Recurrent Neural Networks for Demand Prediction
- Introducing RNNs
- Recurrent Neural Networks
- Recurrent Neural Units
- Long Short-Term Memory
- The Demand Prediction Problem
- Demand Prediction
- Predicting Energy Demand
- Data Preparation – Creating the Past
- Data Loading and Standardization
- Data Cleaning and Partitioning
- Creating the Input Tensors
- Building, Training, and Deploying an LSTM-Based RNN
- Building the LSTM-Based RNN
- Training the LSTM-Based RNN
- Testing the LSTM-Based RNN
- Building a Deployment Loop
- Deploying the LSTM-Based RNN
- Summary
- Questions and Exercises
- Chapter 7: Implementing NLP Applications
- Exploring Text Encoding Techniques for Neural Networks
- Index Encoding
- One-Hot Vector Encoding
- Embeddings for Word Encoding
- Finding the Tone of Your Customers' Voice – Sentiment Analysis
- Preprocessing Movie Reviews
- Defining and Training the Network Architecture
- Executing and Evaluating the Network on the Test Set
- Generating Free Text with RNNs
- The Dataset
- Predicting Words or Characters?
- Preprocessing and Encoding
- Defining and Training the Network Architecture
- Building a Deployment Workflow
- The New Fairy Tale
- Generating Product Names with RNNs
- The Problem of Product Name Generation
- Preprocessing and Encoding Mountain Names
- Defining and Training the Network Architecture
- Building a Deployment Workflow
- Summary
- Questions and Exercises
- Chapter 8: Neural Machine Translation
- Idea of Neural Machine Translation
- Encoder-Decoder Architecture
- Applying the Encoder
- Applying the Decoder during Training
- Applying the Decoder during Deployment
- Preparing the Data for the Two Languages
- Building and Training the Encoder-Decoder Architecture
- Defining the Network Structure
- Training the Network
- Extracting the Trained Encoder and Decoder
- Applying the Trained Network for Neural Machine Translation
- Summary
- Questions and Exercises
- Chapter 9: Convolutional Neural Networks for Image Classification
- Introduction to CNNs
- How are Images Stored?
- Why do we need CNNs?
- How does a Convolution Layer work?
- Introducing Padding
- Introducing Stride and Dilation Rate
- Introducing Pooling
- Classifying Images with CNNs
- Reading and Preprocessing Images
- Designing the Network
- Training and Applying the Network
- Prediction Extraction and Model Evaluation
- Introduction to transfer learning
- Why use Transfer Learning?
- Formal Definition of Transfer Learning
- Applying Transfer Learning
- Applying Transfer Learning for Cancer Type Prediction
- Downloading the Dataset
- Reading and Preprocessing the Images
- Training the Network
- Summary
- Questions and Exercises
- Section 3: Deployment and Productionizing
- Chapter 10: Deploying a Deep Learning Network
- Conversion of the Network Structure
- Saving a Trained Network
- Reading a Trained Network
- Using TensorFlow 2
- Building a Simple Deployment Workflow
- Building a Deployment Workflow Manually, without Integrated Deployment
- Building a Deployment Workflow Automatically with Integrated Deployment
- Improving Scalability – GPU Execution
- Summary
- Questions and Exercises
- Chapter 11: Best Practices and Other Deployment Options
- Building a Web Application
- Introduction to KNIME WebPortal
- Creating a Workflow to Run on KNIME WebPortal
- Creating Composite Views
- Shared Components
- Building a WebPortal Application for Cancer Cell Classification
- Building a Web Service with the REST Interface
- Building a REST Service Workflow
- KNIME Tips and Tricks
- Shuffling Data during Training
- Using Batch Normalization
- Keeping Your Workflow Clean and Structured
- Using the GroupBy Node and Pivoting Node to Avoid Loops
- Specifying the Execution Order
- Summary
- Questions and Exercises
- Other Books You May Enjoy
Section 3: Deployment and Productionizing
Deployment is an important and conclusive phase of a whole data analysis project. So, to conclude, we will show different deployment options for our trained neural networks.
This section comprises the following chapters:
- Chapter 10, Deploying a Deep Learning Network
- Chapter 11, Best Practices and Other Deployment Options
-
No Comment
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.