Deep learning 2019 (half) course

Lectures to be held in MR15, Faculty of Mathematics, Fridays 1-3pm.

Instructors

Stephen J Eglen (Maths), Martin Johnson (AstraZeneca) and Adam Corrigan (AstraZeneca).

Timetable (subject to revision)

1. Feb 22. Introduction: motivating examples. History. From perceptrons to multilayer perceptrons. The importance of features. Practical matters. (SJE) Lecture notes

2. Mar 01. Learning in networks. (Error functions, Back propagation, automatic differentiation, gradient-descent methods). First examples at classification. Hyper-parameters. (SJE) Lecture notes mnist.Rmd mnist.html backprop derivation

3. Mar 08. Image classification and segmentation. Convolutional neural networks. Autoencoders. Applications. (SJE and AC). Lecture notes/1 Lecture notes/2

4. Mar 15. Temporal processing. Recurrent neural networks (RNN) and Long-short-term memories (LSTM). Applications. (SJE and MJ). Lecture notes/1

Text for practical work

Practical work (for the assignment) will be set following the guidance in here. Deep learning with R

Reading

Key referencs will be made available in this Paperpile folder

Assignments

Assignment 1