Cohort 1 student Romana attends the Bumblekite Machine Learning Summer School in Zurich

Romana Burgess had this to say about her experience: This week I attended the Bumblekite Machine Learning Summer School in Healthcare and Biosciences, held at ETH Zurich. The school invited participants from 42 institutions around the world, and a diverse range of speakers from both industry (e.g., Novartis, Microsoft, Genomics England) and academia (e.g., Stanford, ETH, Princeton).

Over the week, I attended many interesting sessions covering a broad range of machine learning topics and applications. For example, we looked at AI techniques for prostate cancer detection, and using fusion models for combining electronic health records and medical images. Additionally, the tutorials provided us with access to some fantastic and unique datasets, allowing us to practice new techniques throughout the week. Each evening I attended an interactive leadership conversation, where academic and industry professionals shared their personal experiences within the ML x Healthcare space. These engaging discussions included ethical considerations in AI and genomics, succeeding in a data science job application, and how to be an inspiring leader.

At the end of the week, we were asked to present on something that we had had learned. My group chose to discuss the benefits and disadvantages of aggregating patient data for use in machine learning models; we discussed loss of information and biased models, versus the importance of patient privacy.

I was fortunate enough that the school organisers gave me the award for “best application” at the end of the school. In my application, I had discussed my desire to learn more about patient care and experience within healthcare systems, and to find new connections. I did in fact build many professional and personal connections with the other applicants over the course of the week; I also spoke with some of the lovely speakers, who gave me great advice and ideas for my own work going forward.

I have always found some more complex machine learning techniques to be intimidating, and I would have been hesitant to try using them myself. However, I have come away from the summer school feeling less apprehensive of these methods, and I believe that some could be worth implementing in my work. For example, I will now consider implementing k-means clustering to identify profiles of depressed or non-depressed parents.