Codeless Deep Learning with KNIME

Build, train, and deploy various deep neural network architectures using KNIME Analytics Platform

Kathrin Melcher, Rosaria Silipo

• Publisher: Packt Publishing
• Year: 2020
• Print ISBN:
  9781800566613
• Edition: 1
• eBook ISBN:
  9781800562424
• Topic: Computer Science
• Subtopic: Computer Vision & Pattern Recognition
• Index: Computer Science

Discover how to integrate KNIME Analytics Platform with deep learning libraries to implement artificial intelligence solutions

Key Features

• Become well-versed with KNIME Analytics Platform to perform codeless deep learning
• Design and build deep learning workflows quickly and more easily using the KNIME GUI
• Discover different deployment options without using a single line of code with KNIME Analytics Platform

Book Description

KNIME Analytics Platform is an open source software used to create and design data science workflows. This book is a comprehensive guide to the KNIME GUI and KNIME deep learning integration, helping you build neural network models without writing any code. It'll guide you in building simple and complex neural networks through practical and creative solutions for solving real-world data problems.

Starting with an introduction to KNIME Analytics Platform, you'll get an overview of simple feed-forward networks for solving simple classification problems on relatively small datasets. You'll then move on to build, train, test, and deploy more complex networks, such as autoencoders, recurrent neural networks (RNNs), long short-term memory (LSTM), and convolutional neural networks (CNNs). In each chapter, depending on the network and use case, you'll learn how to prepare data, encode incoming data, and apply best practices.

By the end of this book, you'll have learned how to design a variety of different neural architectures and will be able to train, test, and deploy the final network.

What you will learn

• Use various common nodes to transform your data into the right structure suitable for training a neural network
• Understand neural network techniques such as loss functions, backpropagation, and hyperparameters
• Prepare and encode data appropriately to feed it into the network
• Build and train a classic feedforward network
• Develop and optimize an autoencoder network for outlier detection
• Implement deep learning networks such as CNNs, RNNs, and LSTM with the help of practical examples
• Deploy a trained deep learning network on real-world data

Who this book is for

This book is for data analysts, data scientists, and deep learning developers who are not well-versed in Python but want to learn how to use KNIME GUI to build, train, test, and deploy neural networks with different architectures. The practical implementations shown in the book do not require coding or any knowledge of dedicated scripts, so you can easily implement your knowledge into practical applications. No prior experience of using KNIME is required to get started with this book.

Section 1: Feedforward Neural Networks and KNIME Deep Learning Extension

This section is introductory. It is here to ease you into the world of neural networks and the tool at hand, KNIME Analytics Platform.

This section comprises the following chapters:

• Chapter 1, Introduction to Deep Learning with KNIME Analytics Platform
• Chapter 2, Data Access and Preprocessing with KNIME Analytics Platform
• Chapter 3, Getting Started with Neural Networks
• Chapter 4, Building and Training a Feedforward Neural Network

Chapter 1: Introduction to Deep Learning with KNIME Analytics Platform

We'll start our journey of exploring Deep Learning (DL) paradigms by looking at KNIME Analytics Platform. If you have always been drawn to neural networks and deep learning architectures and have always thought that the coding part would be an obstacle to you developing a quick learning curve, then this is the book for you.

Deep learning can be quite complex, and we must make sure that the journey is worth the result. Thus, we'll start this chapter by stating, once again, the relevance of deep learning techniques when it comes to successfully implementing applications for data science.

We will continue by providing a quick overview of the tool of choice for this book – KNIME Software – and focus on how it complements both KNIME Analytics Platform and KNIME Server.

The work we'll be doing throughout this book will be implemented in KNIME Analytics Platform, which is open source and available for free. We will dedicate a full section to how to download, install, and use KNIME Analytics Platform, even though more details will be provided in the chapters to follow.

Among the benefits of KNIME Analytics Platform is, of course, its codeless Deep Learning - Keras Integration extension, which we will be making extensive use of throughout this book. In this chapter, we will just focus on the basic concepts and requirements for this KNIME extension.

Finally, we will conclude this chapter by stating the goal and structure of this book. We wanted to give it a practical flavor, so most of the chapters will revolve around a practical case study that includes real-world data. In each chapter, we will take the chance to dig deeper into the required neural architecture, data preparation, deployment, and other aspects necessary to make the case study at hand a success.

In this chapter, we will cover the following topics:

• The Importance of Deep Learning
• Exploring KNIME Software
• Exploring KNIME Analytics Platform
• Installing KNIME Deep Learning – Keras Integration
• Goals and Structure of this Book

We'll start by stating the importance of deep learning when it comes to successful data science applications.

The Importance of Deep Learning

If you have been working in the field of data science – or Artificial Intelligence (AI), as it is called nowadays – for a few years, you might have noticed the recent sudden explosion of scholarly and practitioner articles about successful solutions based on deep learning techniques.

The big breakthrough happened in 2012 when the deep learning-based AlexNet network won the ImageNet challenge by an unprecedented margin. This victory kicked off a surge in the usage of deep learning networks. Since then, these have expanded to many different domains and tasks.

So, what are we referring to exactly when we talk about deep learning? Deep learning covers a subset of Machine Learning (ML) algorithms, most of which stem from neural networks. Deep learning is indeed the modern evolution of traditional neural networks. Apart from the classic feedforward, fully connected, backpropagation-trained, and multilayer perceptron architectures, deeper architectures have been added. Deeper indicates more hidden layers and a few new additional neural paradigms, including Recurrent Neural Networks (RNNs), Long-Short Term Memory (LSTM), Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs), and more.

The recent success of these new types of neural networks is due to several reasons. First, the increased computational power in modern machines has favored the introduction and development of new paradigms and more complex neural architectures. Training a complex neural network in minutes leaves space for more experimentation compared to training the same network for hours or days. Another reason is due to their flexibility. Neural networks are universal function approximators, which means that they can approximate almost anything, provided that their architecture is sufficiently complex.

Having mathematical knowledge of these algorithms, experience with the most effective paradigms and architectures, and domain wisdom are all basic, important, and necessary ingredients for the success of any data science project. However, there are other, more contingent factors – such as ease of learning, speed of prototyping, options for debugging and testing to ensure the correctness of the solution, flexibility to experiment, availability of help from external experts, and automation and security capabilities – that also influence the final result of the project.

In this book, we'll present deep learning solutions that can be implemented with the open source, visual programming-based, free-to-use tool known as KNIME Analytics Platform. The deployment phases for some of these solutions also use a few features provided by KNIME Server.

Next, we will learn about how KNIME Analytics Platform and KNIME Server complement each other, as well as which tasks both should be used for.

Exploring KNIME Software

We will mainly be working with two KNIME products: KNIME Analytics Platform and KNIME Server. KNIME Analytics Platform includes ML and deep learning algorithms and data operations needed for data science projects. KNIME Server, on the other hand, provides the IT infrastructure for easy and secure deployment, as well as model monitoring over time.

We'll concentrate on KNIME Analytics Platform first and provide an overview of what it can accomplish.

KNIME Analytics Platform

KNIME Analytics Platform is an open source piece of software for all your data needs. It is free to download from the KNIME website (https://www.knime.com/downloads) and free to use. It covers all the main data wrangling and m...

Table of contents

• Codeless Deep Learning with KNIME
• Why subscribe?
• Preface
• Section 1: Feedforward Neural Networks and KNIME Deep Learning Extension
• Chapter 1: Introduction to Deep Learning with KNIME Analytics Platform
• Chapter 2: Data Access and Preprocessing with KNIME Analytics Platform
• Chapter 3: Getting Started with Neural Networks
• Chapter 4: Building and Training a Feedforward Neural Network
• Section 2: Deep Learning Networks
• Chapter 5: Autoencoder for Fraud Detection
• Chapter 6: Recurrent Neural Networks for Demand Prediction
• Chapter 7: Implementing NLP Applications
• Chapter 8: Neural Machine Translation
• Chapter 9: Convolutional Neural Networks for Image Classification
• Section 3: Deployment and Productionizing
• Chapter 10: Deploying a Deep Learning Network
• Chapter 11: Best Practices and Other Deployment Options
• Other Books You May Enjoy