Last month, STAI CDT PhD student, Nathan Gavenski, attended the Oxford Machine Learning Summer School (OxML). The summer school focuses on different subjects and Nathan participated in the MLxRepresentation Learning & GenAI track.
The MLxRepresentation Learning & GenAI track explored the latest advances in representation learning, which is the foundation of many recent breakthroughs in generative AI. Sessions covered machine learning across a wide range of domains, including natural language processing and large language models, computer vision with cutting-edge image, video, and multimodal models, Bayesian ML, reinforcement learning, and geometrical deep learning. In addition, several talks highlighted the real-world applications of deep learning and generative AI across key industries and sectors.
The summer school aims to train AI talents, particularly those whose research topics relate to the summer school. This is particularly relevant to Nathan, who attended the summer school because of the focus on Representation Learning, which he plans to use in his thesis. Nathan said, “I believed it would be an ideal setting to not only learn about it but also see how researchers and industry are applying it”.
Nathan particularly enjoyed the opportunity to connect with attendees from across the globe. He said, “The summer school was really nice, they had a good mix of industry and researcher speakers, and the people who attended were from all over the place (Ecuador, China, US, Italy, etc.) and from across different academic and industry levels which led to some interesting conversations and networking”. He also enjoyed being able to explore and get to know Oxford.
When asked what his highlights were, Nathan said, “I’m a little biased because some talks were on imitation learning (my research topic), and it is pretty rare to see people talk about it. For me, the main highlight was a talk from Haitham Bou-Ammar (UCL) that drew inspiration from learning theory, a subject I enjoy and is a part of my research. He talked about his research with Huawei and his lab at UCL, and how they apply Kolb’s experiential learning theory to develop learning agents for tasks in robotics and programming. It is nice to see approaches that are human-inspired in computer science”.
Nathan found the summer school really inspiring and taking part in it has also led to lots of new ideas for Nathan’s own thesis. For Nathan, a key takeaway was recognising the value of attending events like this, having previously focused more on getting published and attending conferences. He particularly found it useful to see how other researchers who are working in a similar way to him (there is a big component to Behaviour Science in his PhD) are looking to other fields, such as robotics, to find more use cases of imitation learning.