Overview
The Machine Learning (ML) stream of the MPhil in Scientific Computing brings together the elements needed to make sense of scientific data. It is used in many disciplines such as Physics, Chemistry, Astronomy, Financial Mathematics, Medicine, Bioinformatics and Robotics. The problems ML endeavours to solve are often put together under the term Big Data describing any large amount of structured, semistructured and unstructured data. The techniques are widely used, for example in fraud detection, search engines, gaming and advertising.
Prerequisites
Students following this stream should have a first degree in a numeric discipline such as mathematics, physics or engineering. An understanding of probability theory and statistics is a plus, but not essential, as is experience in a high level programming language.
Recommended reading
Christopher M. Bishop "Pattern Recognition and Machine Learning", Springer, 2007.
Kevin P. Murphy "Machine Learning: a Probabilistic Perspective", MIT Press, 2012.
David J.C. MacKay "Information Theory, Inference, and Learning Algorithms", Cambridge University Press, 2003.
David Barber "Bayesian Reasoning and Machine Learning", Cambridge University Press, 2012.
Lecture courses
Typically, students who are part of this stream are expected to attend the following lecture courses:
 Machine Learning
 Linear Systems
 Fundamentals, Nonlinear Systems, Interpolation
 Introduction to Linux
 Scientific Programming in C++
 Software Design
 GPUs for Scientific Computing
 Message Passing Interface (MPI)
 OpenMP
 Computational Hardware
 Programming for Powerefficient Computing
Examinations
Typically, students who are part of this stream are expected to sit the following examination papers:
Paper 1: Fundamentals in Numerical Analysis (12hrs, 2units).
Paper 4: Linear Systems (12hrs, 2units).
Paper 7: Machine Learning (12 hours, 2 units).
Research Projects
The 201516 cohort investigated the following topics:

On the Analysis of HTS Data from SELEX Experiments

Compressive Sensing in Video Reconstruction

Detection Techniques for Drone SenseandAvoid Radar

The Model is Simple until Proven Otherwise – How to Cope in an Ever Changing World

Valuing Multidimensional Bermudan Put and Call Options

Mining Vehicle Electronic Control Unit (ECU) Data for Driver and Drive Cycle Identification

A MetaAlgorithm for Classification using Random Recursive Tree Ensembles: A High Energy Physics Application

GraphBased Clustering: Distributed Algorithms for Balanced Graph Cuts

Sequential Matrix Completion

Deep Learning to Exploit MultiSpectral Satellite Data for Landcover Mapping and Crop Classification