Skip to 0 minutes and 2 seconds Welcome to the introductory lecture of Advanced Machine Learning with FutureLearn, Open University, and Persontyle. So I’m going to be your lecturer guiding you through this course. My name’s Michael Ashcroft. I’m a professor over at Uppsala University. And I’m also involved in all sorts of projects within industry and commerce including such things as financial technology, security and counterterrorism, industrial diagnostics with some of Europe’s logic companies, social media analytics, et cetera, et cetera. And there’s all sorts of things that machine learning is used for these days. And I’ve had a go at a lot of them. As well as dividing the sort of projects I’ve worked on in these sorts of areas, I’ve also worked on all sorts of data.
Skip to 0 minutes and 54 seconds We look at video data, text data, and your basic structured data of the type you think of when you think of a Excel spreadsheet or a relational database. Also of course, things like real time and static data projects. So I’ve got a lot of experience actually applying these things in the field, as well as lecturing on them at Uppsala University in Sweden, and working through a lot of projects with master students. Now this course is designed to go a little bit deeper than a lot of machine learning and AI courses that are out there.
Skip to 1 minute and 32 seconds We’re going to start off by looking at the theory in the background, the sort of things that you need to know in order to go deeper. We will then consider how we can analyse statistical models, where statistical models of the sorts of things that we’re going to be getting out of our machine learning algorithms. And how to analyse the data that we use to generate these statistical models.
Skip to 1 minute and 58 seconds So the theory and background will take most of first week. Analysis of models and analysis of data will be the start of the second week. And then we’ll start looking at some of the more sophisticated modelling techniques. We’re going to start off with supervised learning, which we’ll talk about what that is soon in the course. The models we’ll be looking at will be things like artificial neural networks, which of course, very popular these days with deep learning. Now we’re not going to go to deep learning, that’s a separate course. But we are going to give you everything you need to know to understand deep learning and understand artificial neural networks in general.
Skip to 2 minutes and 38 seconds Then we’ll also look at a whole collection of methods called kernel methods. After these supervised learning techniques at the beginning of week three, we will be looking at unsupervised learning, looking at things like your clustering methods, LDA for topic analysis of documents. Finally, we’ll finish up by looking at some feature engineering, how you can transform the data that you’re given into more informative data essentially. And also, how we can deal with missing data. So in a way, we go theory, analysis of the models and the data, bunch of modelling algorithms, and then some pre-processing stuff, how we can get the data in the right sort of shape to optimally work with the models we’ll look at.
Skip to 3 minutes and 29 seconds So for you who are taking this course, as I said, I wanted to give you a relatively deep understanding of the methods we’re going to look at, to go a little bit further than the introductory courses that are available around the place. We’re going to give you rigorous methods for analysing the sort of things that you’re doing, the data you’re working with, the models you’re generating, et cetera. And we’re going to give you a really good basis of machine learning theory. Now in this first week as I said, we’ll look at background theory and knowledge.
Skip to 4 minutes and 1 second This is going to include having a look at the data science workflow, what is involved in typical data science project, and in particular, which steps are those that a machine learning expert, as opposed to a general data scientist, would really be involved in. Then we’ll start talking about feature transformations and how you can generate non-linear models from linear models using them. We’ll talk about analysis of error, which is the bias variance decomposition, which you may know about. And we’ll look at that in a little bit of detail. We’ll look at learning a statistical model as a process of optimization, as an optimization problem.
Skip to 4 minutes and 44 seconds And then we’ll discuss regularisation, how we can resist allowing sophisticated models to overfit on a given set of data. So I’m really excited about giving this course. And I hope you’re all going to enjoy it. And let’s get stuck into it. Thank you.
Introduction to Course & Week One
A short introduction video to the course and an introduction/background to the lecturer Mike Ashcroft, as well as an overview of the topics we will look at in week one.
© Dr Michael Ashcroft