X Close

Science blog


News, anecdotes and pictures from across science and engineering at UCL


UCL STFC summer school in Machine Learning and Artificial Intelligence

By Charlotte E Choudhry, on 26 September 2018

This July, University College London (UCL) hosted the first ever Science and Technology Facilities Council (STFC) summer school in Artificial Intelligence (AI) and Machine Learning.
120 PhD students from universities across the country were brought together to spend a week and a half learning about the forefront of AI technology. With lectures from several industry leaders; including Intel, NVidea, ASI, and Dell, the summer school aimed to give a solid grounding in the basics of machine learning before providing a glimpse into some of the latest technologies now being applied.

This included being granted access to some of the best interactive platforms currently available; and allowed students to follow along with demonstrations during the lectures. To start with, ASI’s ‘SherlockML’ enabled students to familiarise themselves with the most common tools used in machine learning. By making use of the Intel ‘DevCloud’, students were then shown how to optimize Neural Networks in TensorFlow and Caffee, before learning about computer vision and making use of the Movidius stick to be able to turn a webcam into a tool to classify everyday objects. NVidea also demonstrated how to make the most out of currently available computing clusters, using Graphics Processing Units (GPUs) to parallel process tasks, and students were taught how to build a convolutional neural network in ‘Digits’ which could then classify whale faces.

On top of the extensive technical training there were additional guest lectures highlighting how machine learning can be applied to a variety of different fields, from high energy physics and astrophysics to in industry. It was also made evident how machine learning techniques will be vital in the future of astronomy with new methods such as gravitational waves and new observatories like the Square Kilometer Array (SKA) which will produce petabytes of data every day.
Over the weekend a further 100 sixth form pupils were invited to take part in a hackathon-style event lead by the visiting PhD students, where during the afternoon the students demonstrated how machine learning could be applied in a real research scenario at the Large Hadron Collider (LHC).

One of the best things about assembling people from such a wide range of disciplines, with a common interest in machine learning and AI, was the incredible diversity of the work presented. Over the course of three poster sessions, we had the opportunity to discover what was being researched; from classifying solar winds and galaxies to using neural networks for novel detection methods in the LHC. And for one evening, this was combined with a chance to meet and talk with professionals from various CDT industry partners with companies like ASOS, TFL, and NCC Group.

As well as networking with the industry partners, a key aim of any summer school is to develop links between the students themselves. Although never an issue when helped along with some free food and drinks, it was especially encouraging to see everyone come together for the quiz night, with each team setting a round of obscure questions, and the school ended in spectacular fashion making use of the record-breaking summer with a river cruise BBQ on the Thames.

The Summer School was directed by Jonathan Tennyson, and the LOC/SOC included Tim Scanlon & Jason McEwen (CDT Research Directors), Jaini Shah (CDT Manager),  and Nikos Konstantinidis & Ofer Lahav (CDT co-Directors).

-Written by Ben Henghes (Research Student, UCL Physics & Astronomy)

2 Responses to “UCL STFC summer school in Machine Learning and Artificial Intelligence”

  • 1
    Jeetech Academy wrote on 26 August 2021:

    I really enjoyed this article.
    Great tips to thank you … keep updating

  • 2
    Jeetech Academy wrote on 20 September 2021:

    You have done a fantastic job. Thank you for all the details.

Leave a Reply