PLEASE DO NOT REGISTER FOR THIS EVENT IF YOU HAVE NO INTENTION OF ATTENDING.

Although places are free to students, the Q-Step Centre incurs substantial costs for these events which cannot be refunded in the event that registered students do not turn up on the day. Please make every effort to inform us in advance of the event if you are unable to attend or you can cancel your place directly on Eventbrite.

The Warwick Q-Step Centre is delighted to host this Masterclass on Deep Learning, delivered by Dr Christian Arnold - Lecturer in Politics at Cardiff University.

This Masterclass is aimed primarily at postgraduate students on our 3 Quantitative Social Science degrees. Final year Q-Step undergraduate students, interested in learning enhancement opportunities, are also invited to attend. Depending on numbers we may be able to open the event up to PG students in other relevant subject areas throughout the University.

Deep Learning

Neural networks – long believed dead – are back. Deep Learning has revolutionised Computer Science and machine learning a decade ago and is continuing to build the foundation of powerful modern data science applications.

The reason behind this recent surge is quite straightforward. As always, ever bigger and faster computers matter. In addition, digitalisation produces massive amounts of data. Deep neural network architectures excel at this particular challenge, since they require truly large amounts of data to work well. The advent of deep neural architectures promises unforeseen advances in prediction. No wonder both scientists but also data intense businesses are fast in adopting these techniques.

This class will provide an introduction to deep neural networks. Starting from flat architectures, we will understand how to use backward propagation to estimate neural networks. Adding more layers to the initial set-up, we will then go on and build increasingly “deep” architectures. We implement fully connected deep learning architectures and learn about classical pitfalls and how to tackle them. If time permits, we will give up the full connectedness between nodes and take a first look at models that further aggregate information to more abstract data representations.

Computing Requirements

For teaching, the class will alternate between lecture style blocks and more hands-on workshop style blocks where participants implement code themselves. A range of different languages exist to build deep learning models. This class relies on R and KERAS as main APIs. Please make sure you have the latest release of R, RStudio and KERAS installed on your machines when you come to class. Please do not forget your charging cables: high performance computing is a drain on energy and you will not be able to follow the workshops on battery, only.

Literature

The class has a main textbook it draws from:

Chollet, François and J.J.Allaire. 2018. Deep Learning with R,Manning: Shelter Island.

Deep Learning is a quickly developing field. In addition to the book, the internet provides a host of excellent sources that keep track of recent developments. I can for example recommend the following github-page that keeps track of the most relevant papers since 2012.

This book is more of a deep dive into it all and recommended for further reading.

Goodfellow, Ian, Yoshua Bengio and Aaron Courville. 2016. Deep Learning. Cambridge University Press.

PLEASE NOTE: This event is for Warwick University students only. There is no option to attend parts of the Masterclass, you must be able to attend for the whole day.

Attendance is free and refreshments, including lunch, will be provided. Registration is required as places are strictly limited. So book your place now!

Further information about the Warwick Q-Step Masterclass Programme 2019/20 can be found on our website.