CDT Digital Health and Care Cohort 1 student Romana Burgess reports back from the “MHINT micro-coding workshop” that she recently participated in:

On Friday 2nd July, I attended a virtual workshop to discuss micro-coded videos of parent-child interactions, with the aim to foster collaborations moving forward in how to analyse these complicated datasets.

The event joined together 20 academics from multiple countries (e.g. UK, Brazil, Chile, Norway, South Africa), disciplines (e.g. epidemiologists, mathematicians, dieticians, psychologists, biologists), and career levels (e.g. professors, lecturers, post docs, PhD and MSc Students).

Many of these researchers have collected and processed their data, and are now considering how to analyse it. This is a core focus of my PhD, and so I – along with multiple other academics – presented some ideas for potential next stages. For example, how we might use a facial analysis software to automatically capture and describe facial expressions, or how we might identify sequences of behaviours from coded data.

Not only was the workshop beneficial for developing collaborative projects, but it also provided me with an opportunity to practice communicating my research to other academics. Particularly, I was able to practice explaining mathematical and computing-based concepts to those within non-STEM backgrounds.

Harry says of the event:

The summer school was a week-long introduction to the theory and implementation of reinforcement learning algorithms, organised by the Alan Turing institute. Each day commenced with a series of online lectures delivered by Dr Nathanaël Fijalkow explaining the mathematics underlying a selection of significant reinforcement learning algorithms. The topics covered were broad, allowing me to learn about a variety of machine learning and deep learning techniques, including Monte Carlo Tree Search, Deep Q Learning and Policy gradient methods. This knowledge was supplemented with daily workshops, where students were divided into structured seminar groups and asked to collaboratively work towards implementing the algorithms described in the morning lecture. This provided me with an opportunity to get feedback on my implementations from experts in the field, as well discuss my research and ideas with other students starting their own research in reinforcement learning.

I am really pleased I had the opportunity to attend the summer school, as it is highly relevant to my PhD project. In the coming years, I will focus on implementing reinforcement learning algorithms to automate the control of insulin delivery for type 1 diabetics. Applying this technique to medical data comes with a series of associated challenges around utilising small and often sparse datasets, minimising patient risk and evaluating algorithmic performance. Having a good foundational understanding of the inner workings and limitations of reinforcement learning algorithms will make me better equipped to overcome these obstacles in my project. The course also gave me the opportunity to discuss my research with individuals working in a diverse range of fields, such as environmental data science and astrophysics. Listening to the successes and difficulties they have had in overcoming the challenges in their research, also provided me with inspiration as to the direction I could go in my own project.

From the 25-29 Jan 2021 Cohort 1 student Holly Fraser attended the Psychological Networks Amsterdam Winter School 2021. The winter school provided training in Network Analysis, a methodology used to visualise and explore associations between phenomena at a systems-level. Participants learnt how to implement network estimation algorithms from the leading academics in the field, understand how to map out and infer causal relationships between factors, and learn about the future directions of Network Analysis. The course was delivered via asynchronous learning, with the course hosts and academics on hand via a Slack channel to answer any questions and help through the exercises.

Holly says: ‘I attended to learn more about the method so I can apply it to my own work in the future. I did some of my own data analysis alongside the taught activities, and I now feel much more confident to share my findings and results with my research group and plan future publications. I anticipate that I’ll use this method in the initial stages of my PhD, as a means of data exploration and hypothesis generation. The method is a great way to visualise data, so exploring the ALSPAC dataset through network estimation will inform my later machine learning analyses. As this method hasn’t been implemented before in this dataset, there is potential for some exciting future work to arise from this analysis, such as understanding connections between mental health symptoms.  I hope to also collaborate with colleagues from other universities who are interested in using networks to explore mental health