In this two day course, we provide an introduction to machine learning and deep learning using R. We begin by providing an overview of the machine learning and deep learning landscape, and then turn to some major machine learning applications. We begin with binary and multiclass classification problems, then look at decision trees and random forests, then look at unsupervised learning methods, all of which are major topics in machine learning. We then cover artificial neural networks and deep learning. For this, we will use the Torch deep learning library. As examples of deep learning nets, we will cover the relatively easy to understand multilayer perceptron and then turn to convolutional neural networks.

Intended Audience

This course is aimed at anyone who is interested in learning the machine learning or deep learning using R, both of which have major applications both in industry and in academia.

Teaching Format

This course will be hands-on and workshop based. Throughout each day, there will be some brief introductory remarks for each new topic, introducing and explaining key concepts.

The course will take place online using Zoom. On each day, the live video broadcasts will occur between (UK local time) at:

• 10am-12pm
• 1pm-3pm
• 4pm-6pm

All sessions will be video recorded and made available to all attendees as soon as possible, hopefully soon after each 2hr session.

Attendees in different time zones will be able to join in to some of these live broadcasts, even if all of them are not convenient times.

By joining any live sessions that are possible, this will allow attendees to benefit from asking questions and having discussions, rather than just watching prerecorded sessions.

Although not strictly required, using a large monitor or preferably even a second monitor will make the learning experience better.

All the sessions will be video recorded, and made available immediately on a private video hosting website. Any materials, such as slides, data sets, etc., will be shared via GitHub.

Assumed quantitative knowledge

We will assume familiarity with some general statistical and mathematical concepts such as matrix algebra, calculus, probability distributions. However, expertise with these concepts are not necessary. Anyone who has taken any undergraduate (Bachelor’s) level course in mathematics, or even advanced high school level, can be assumed to have sufficient familiarity with these concepts.

Assumed computer background

We assume general familiarity with using R and RStudio, and some familiarity with programming in R, such as writing code.

Equipment and software requirements

Attendees of the course must use a computer with R/RStudio installed, as well as the necessary additional R packages.

Course programme

Day 1

• Topic 1: Machine learning and Deep Learning Landscape. Concepts like machine learning, deep learning, big data, data science have become widely used and celebrated in the last 10 years. However, their definitions are relatively nebulous, and how they related to one another and to major fields like artificial intelligence and general statistics are not simple matters. In this introduction, we briefly overview the field of machine learning and deep learning, discussing its major characteristics and sub-topics.

• Topic 2: Classification problems. Classification problems is one of the bread and butter topics in machine learning, and is usually the first topic covered in introductions to machine learning. Here, we will cover image classification (itself a “hello world” type problem in machine learning), and particularly focus on logistic regression and support vector machines (SVMs).

• Topic 3: Decision trees and random forests. Decision trees are a widely
  used machine learning method, particularly for classification. Random forests
  are an ensemble learning extension of decision trees whereby large number of decision tree classifiers are aggregated, which often leads to much improved performance.

Day 2

• Topic 4: Unsupervised machine learning. Unsupervised learning is essentially a method of finding patterns in unclassified data. Here, we will look at two of the most widely used unsupervised techniques: k-means clustering and probabilistic mixture models.

• Topic 5: Introducing artificial neural networks and deep learning. R provides many packages for artificial neural networks and deep learning. These include Keras and TensorFlow, which are in fact interfaces to Python packages. These are the most widely used major packages for deep learning in R. More recently, native support for deep learning using R via Torch has been introduced, which we use in this course.

• Topic 6: Multilayer perceptons. Multilayer perceptrons are very powerful, yet relatively easy to understand, artificial neural networks. They are also the simplest type of deep learning model. Here, we will build and train a multilayer perceptron for a classification problem.

• Topic 7: Convolutional neural networks. Convolutional neural networks (CNNs) have proved high successful at image classification, primarily due to their translation invariance, which is highly suitable for computational vision. Here, we use PyTorch to build and train a CNN for image classification.

GitHub resources

Further resources for this training course can be found on Github at mark-andrews/imdl02.