Summer School on Machine Learning in Bioinformatics Held at HSE University
The second international Summer School on Machine Learning in Bioinformatics took place on August 23–27. This year’s school featured 533 participants from 53 countries.
Among them were students and graduates of HSE University, Skoltech, MIPT, Moscow State University (Russia), École Polytechnique (France), Heidelberg University (Germany), Sapienza Università di Roma, The University of Bologna (Italy), MIT, Northeastern University, The University of California San Diego (US), The University of Oxford, The University of Cambridge, Imperial College London, The University of Sussex, The University of Glasgow (UK) and many others.
Topics of the Summer School included machine learning applications in genetics and proteomics, graph neural networks for protein 3D structures, and bayesian checkpoint segmentation and genome annotation. The list of speakers included Maria Poptsova and Nazar Beknazarov (HSE University), José Miguel Hernández-Lobato (University of Cambridge), Daisuke Kihara (Purdue University), Matthias Heinig (Helmholtz Zentrum München), Wesley De Neve (Ghent University), Asa Ben-Hur (Colorado State University), and Yaron Orenstein (Ben-Gurion
University of the Negev).
The participants and organisers shared their thoughts on the school:
Laboratory Head, International Laboratory of Bioinformatics
I was surprised at how many active scientists from all over the world were speakers at the school. They talked about their latest research and shared practical skills, and it was a learning experience for me as well as the students. One of the clear benefits of online conferences is that they don’t just remove the barriers between countries—they remove the barriers of time and space!
Ėcole Polytechnique, France
I’ve really enjoyed the Summer School so far. The lectures are diverse and very interesting. They perfectly suited my goal of learning more about the use of machine learning in Biology, and I’ve been able to learn quite a lot of state-of-the-art things. The format is a bit intense, but knowing the topics in advance allowed me to prepare for the talks I wanted to take good notes on, as well as which ones I could listen to in order to satisfy my curiosity. I’ve also really enjoyed the partial sessions, which give a more concrete understanding of the somewhat abstract concepts discussed in the lectures, and provide a bare-bones code template if I ever need to use these concepts again. As an incoming graduate student planning to study artificial intelligence in genomics, it’s useful to have a good stepping-off point for all of these topics.
As for the talks that interested me the most, I would have to say that Manifold Learning and the Interpretability of Neural Networks were the most interesting. I’ve already worked a bit with Deep Neural Networks in Genomics, so I appreciated learning about what exists beyond the standard 1D convolutional neural networks (CNNs). The refresher lectures were also very helpful, as they let me confirm that my knowledge of CNNs was solid, and even reminded me of some technical details I had forgotten.
All in all, I think your lecturers provide a good mix of topics and difficulty levels for all attendees.
University of California, San Diego, USA
In terms of the knowledge gained, because I have more of a biological background, the majority of the talks were very advanced for me. However, I found it really useful in terms of making a ‘reading list’ for future topics to learn about. The topic I found personally most interesting was explainability in neural networks—finding out how to identify and remove artefacts is extremely important in the work I plan on doing, as there are lots of confounding factors when investigating genetic causes of substance abuse and mental illness. Overall, I found the summer school very useful and definitely worth staying up for.
University of Milan-Bicocca, Italy
As an undergraduate student of a pure biology degree with little to no experience in machine learning and bioinformatics, I am finding myself struggling a bit to keep up due to my own limitations. However, the speakers are absolutely clear and engaging, and they help me to understand despite my ignorance. One of the aspects of this course that impressed me the most is how connected I feel both to the other students and to the professors: we share resources through a Telegram group chat and lessons are not in the classic teacher-fronted format. They feel more like an open discussion, given how many interesting questions keep coming up. Overall, I believe this course is extremely valuable due to the interesting topics it covers and the network that students can create, and I also think that what I'm learning will be useful in my further studies in bioinformatics.
Vectorspace AI, Australia
I was very impressed by the bioinformatics and machine learning summer school run by HSE University. Professors from all around the world provided very interesting and in-depth educational content on these topics, the problems that exist today, and how they are trying to solve them.
The Google Colabs were very engaging and allowed me to get my hands on some working machine-learning code during the seminars. I was able to talk to, and network with, many interesting people before, during and after the sessions.
The course has further ignited my passion for bioinformatics and machine learning and helped me to identify many areas in which I need to improve my understanding.
I hope HSE University will host another summer school next year!