Neural networks and Keras

This chapter discusses the basic operation of an artificial neural network which is the major paradigm of deep learning. The name derives from an analogy to a biological brain. The discussion begins by outlining the basic operations of neurons in the brain and how these operations are abstracted by simple neuron models. It then builds networks of artificial neurons that constitute much of the recent success of AI. The focus of this chapter is on using such techniques, with subsequent consideration of their theoretical embedding.

Basic probability theory

The discussion provides a refresher of probability theory, in particular with respect to the formulations that build the theoretical language of modern machine learning. Probability theory is the formalism of random numbers, and this chapter outlines what these are and how they are characterized by probability density or probability mass functions. How such functions have traditionally been characterized is covered, and a review of how to work with such mathematical objects such as transforming density functions and how to measure differences between density function is presented. Definitions and basic operations with multiple random variables, including the Bayes law, are covered. The chapter ends with an outline of some important approximation techniques of so-called Monte Carlo methods.

Scientific programming with Python

This chapter offers a brief introduction to scientific programming with Python with an emphasis on some mathematical operations that will form the basis of many algorithms. This will specifically include working with matrices and convolutions. Python is a high-level programming language similar to Matlab and R that has gained increasing popularity in the machine learning community. The main reason this book uses Python is that it is freely available and now provides considerable support for machine learning with packages such as sklearn and Keras that are discussed and utilized in this book. Some familiarity with programming concepts is assumed, and the chapter concentrates on a brief introduction to the specific environment and supporting libraries used throughout as well as some basic operations such as convolutions that will be important in later algorithms.

Artificial intelligence, the brain, and our society

The concluding chapter is a brief venture into a more general discussion of machine learning, how it relates to artificial intelligence (AI), and the recent impact of this on society. It starts by discussing the relations of machine learning models in relation to the brain and human intelligence. The discussion then moves to the relation between machine learning and AI. While they are now often equated, it is useful to highlight some possible sources of misconceptions. It closes with some brief thought on the impact of machine learning technology our society.

Cyclic models and recurrent neural networks

This chapter discusses models with cyclic dependencies. There are two principle architectures that are discussed. The first principle architecture of cyclic graphs comprises directed graphs similar to the Bayesian networks except that they include loops. Formally, such networks represent dynamical systems in the wider context and therefore represent some form of temporal modeling. The second type of models have connections between neurons that are bi-directional. These types of networks will be discussed in the context of stochastic units in the second half of this chapter.

Generative models

This chapter presents an introduction to the important topic of building generative models. These are models that are aimed to understand the variety of a class such as cars or trees. A generative mode should be able to generate feature vectors for instances of the class they represent, and such models should therefore be able to characterize the class with all its variations. The subject is discussed both in a Bayesian and in a deep learning context, and also within a supervised and unsupervised context. This area is related to important algorithms such as k-means clustering, expectation maximization (EM), naïve Bayes, generative adversarial networks (GANs), and variational autoencoders (VAE).

Machine learning with sklearn

This chapter’s goal is to show how to apply machine learning algorithms in a general setting using some classic methods. In particular, it demonstrates how to apply three important machine learning algorithms, a support vector classifier (SVC), a random forest classifier (RFC), and a multilayer perceptron (MLP). While many of the methods studied later go beyond these now classic methods, this does not mean that these methods are obsolete. Also, the algorithms discussed here provide some form of baseline to discuss advanced methods like probabilistic reasoning and deep learning. The aim here is to demonstrate that applying machine learning methods based on machine learning libraries is not very difficult. It offers an opportunity to discuss evaluation techniques that are very important in practice.

Reinforcement learning

The discussion here considers a much more common learning condition where an agent, such as a human or a robot, has to learn to make decisions in the environment from simple feedback. Such feedback is provided only after periods of actions in the form of reward or punishment without detailing which of the actions has contributed to the outcome. This type of learning scenario is called reinforcement learning. This learning problem is formalized in a Markov decision-making process with a variety of related algorithms. The second part of this chapter will use function approximators with neural networks which have made recent progress as deep reinforcement learning.

Regression and optimization

This chapter returns to the more theoretical embedding of machine learning in regression. Prior chapters have shown that writing machine learning programs is easy using high-level computer languages and with the help of good machine learning libraries. However, applying such algorithms appropriately with superior performance requires considerable experience and a deeper knowledge of the underlying ideas and algorithms. This chapter takes a step back to consider basic regression in more detail, which in turn will form the foundation for discussing probabilistic models in following chapters. This includes the important discussion of gradient descent as a learning algorithm.

Introduction

This chapter provides a high-level overview of machine learning, in particular how it is related to building models from data. It starts with placing the basic concept in its historical context and phrases the learning problem in a simple mathematical term as function approximation as well as in a probabilistic context. In contrast to more traditional models, machine learning can be characterized as non-linear regression in high-dimensional spaces. This chapter points out how diverse subareas such as deep learning and Bayesian networks fit into the scheme of things and motivates further study with some examples of recent progress.