Machine Learning for particle physics

Brendan Martin is currently in year 4 of the MPhys Mathematical Physics degree.  He completed a Career Development Summer Project in machine learning.

I worked in the area of machine learning for particle physics. Machine Learning can be an extremely useful tool for analysing data from experiments – in classifying particles or identifying interesting event topologies, for example. Designing accurate, computationally cheap algorithms is therefore hugely important. Under the supervision of Prof Luigi Del Debbio, I investigated the relationship between the bias, variance and noise of a given data set using a deep neural network as an estimator. I gained insight into the fascinating, quickly developing field of machine learning whilst simultaneously improving my programming skills.

I enjoyed the freedom that I was afforded during the project; I was able to look into a whole host of different machine learning algorithms, such as convolutional neural networks and recurrent neural networks. Furthermore, I realised the importance of regularisation and learned about different ways to regularise a learning algorithm.

It was exciting to work on an extensive problem where the outcome was not known from the outset. The experience confirmed my enthusiasm for original research and motivates me to continue contributing to the physics community.

Learn more about the School of Physics and Astronomy Career Development Summer Projects.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.