April 1, 2021, by Prof Meghan Gray

First graduates from the MSc Machine Learning in Science

This is a long-overdue post to mark the December 2020 graduation of the first cohort from the one-year postgraduate MSc Machine Learning in Science degree offered by the School of Physics and Astronomy. The field of machine learning and artificial intelligence is growing rapidly and these tools are being exploited by researchers in the School of Physics and astronomy working on fields from nanoscience to astronomy.

Students on the MSc are drawn from backgrounds in physics, mathematics, computer science, chemistry, engineering and learn how to apply ML and AI techniques to real scientific problems. This helps build vital skills, enhancing employability in a rapidly expanding area.

And while the first COVID lockdown interfered with the much anticipated final self-driving car challenge, two of our recent MSc graduates, Luke and Sunny, have kindly provided the following reflection on their experience on the course.

Luke Camilleri (BE Electrical and Electronics Engineering, University of Malta 2019):

Machine Learning has been one of the emergent technologies of the past decade. The rapid development of the field has led to impressive technological break throughs in speech recognition such as Siri and Alexa, driverless cars and face recognition the name a few. Studying machine learning has become something of a cliché lately. However, it is an invaluable tool that can be applied to many different fields of study and can help complement scientific research and development. As an electrical & electronics engineer, my interest in how machine learning algorithms could complement more traditional engineering solutions led me to pursue an MSc in Machine Learning in Science at the University of Nottingham.

During this Master’s course I expanded my knowledge of machine learning principles and their underlining statistical foundations. The focus on assignment-based assessment allowed me to develop practical machine learnings skills. It also gave me the breathing room I needed to understand more complex theoretical concepts.

I also had the opportunity to work with Onyx InSight on my dissertation project. Onyx InSight is an engineering company that helps wind turbine operators reduce their operations and maintenance costs through predictive maintenance and the analysis of vibration signals. The goal of my project was to explore the potential of machine learning to detect patterns in vibration data to aid in the automatic detection and classification of various wind turbine component faults without the intervention of human experts. Such a solution would allow for wide range deployment as the inference process could become automated. Working with Onyx InSight exposed me to the challange of working with real world data and developing real world machine learning solutions. Dealing with problems such as data quality, inconsistent labels and large datasets has provided me with invaluable skills that are already serving me well in my career as a machine learning engineer.

Reading for my Machine Learning in Science MSc and studying at the University of Nottingham was a thoroughly enjoyable experience that has helped me develop on a personal and academic level. The skills learnt through the Masters programme have opened up new and exciting opportunities that I hope to develop in my career and further studies.

Sunny Howard (BSc Physics, University of Nottingham 2019):

As the end of my Physics BSc beckoned, I began to ask myself about what I should do next. I had found the computational parts of the degree particularly fulfilling, so I set about finding myself a MSc programme in a relevant field. In this research I discovered machine learning, and the way that it had revolutionized the world. The power of the techniques to solve previously unsolvable problems fascinated me, and I knew that by taking a course in this I would be highly employable. Considering I had already spent 3 years at UoN, I was aware that I am very happy here, and I think one would struggle to find a student who didn’t enjoy their time in Nottingham. These factors caused me to enrol on the Machine Learning in Science MSc. One further thing to mention is that during my BSc I took a module named Symmetry and Action Principles and found that Professor Juan Garrahan was one of the best teachers I had been taught by. The fact that he was acting as one of the course conveners gave me confidence.

The course covers a great deal of content starting from the first principles of machine learning. For me this was essential, as I believe that it is very important to understand the techniques that you are implementing. If you do not, then when they do something unpredictable (which happens often in machine learning), you will have no way of determining the problem. Further to this, a fundamental understanding allows you to develop existing methods as well as creating new ones. We also had the choice of optional modules that covered a wide range of science, allowing us to apply some of the techniques that we had learned. Some of the modules that I chose were Computational Neuroscience and The Physics of Deep Learning. The latter was particularly interesting and allowed me to bridge the gap between my undergraduate and postgraduate courses, by realising the relation between the Ising model (a model of atomic spins in statistical physics) and the Hopfield model (an associative memory model in machine learning).

The degree culminates in a final project, and we were offered several set possibilities but also the option to find a project supervisor from anywhere in the university (if the project was suitable). I took a project with Dr Edward Gillman from the department of Physics, in which we developed a new method for performing reinforcement learning using tensor networks. This project was exactly what I was looking for as it involved creating a new machine learning method, rather than applying an existing one to a scientific problem. The project went well, and we even had enough time to extend the model to multi-agent reinforcement learning.

I am currently a PhD student at the University of Cambridge in the department of Material Science. The aim of my research is to aid the development of memresistive devices, which have the potential to revolutionize random-access memories (RAM) in computers, reducing energy consumption by over 50%. The coding skills that I developed during my MSc have proven essential in the computational projects I have undertaken, involving analytical modelling and density functional theory (DFT) simulations. Although I have not yet used machine learning in my research, I have already noted several opportunities where it may prove useful, such as in finding suitable functionals in DFT. I am highly confident that by the end of my PhD I will have implemented machine learning algorithms.

Congratulations to Sunny and Luke and all the other MSc graduates!

• Share on Facebook
• Share on Twitter
• Share on Google+
• Share on LinkedIn