Arrykrishna Mootoovaloo

Virtual Workshops in 2020

Highlights from ICLR, ICML, MLSS, OxML and GPSS, all online in 2020.

1 min read Machine Learning and Statistics

In 2020, the COVID-19 pandemic moved most conferences and schools online, and working from home became the norm. The virtual format made it possible to attend more events than would otherwise have been practical. In this post, I summarise the virtual workshops and schools I attended that year.

  1. ICLR 2020 (26th April to 1st May, )

    2. I had the opportunity to present my work on weak lensing, data compression and Gaussian Processes (see here) at ICLR 2020. Beyond my own work, it was inspiring to see how machine learning has grown from a purely scientific field into an engineering discipline that now spans almost every branch of science and engineering. One of the main themes was climate change, one of the most pressing challenges facing society.

  2. MLSS 2020 (28th June to 10th July)

    3. Although I did not register for this school, all talks were streamed live on YouTube. As my work is closely related to Bayesian analysis, I found the talks by Shakir Mohamed particularly valuable (Bayesian Inference I and II). Another notable talk was on meta-learning by Prof. Yee Whye Teh, in which he quoted the following:

    "Our training procedure is based on a simple machine learning principle: test and train conditions must match" - Vinyals et al. 2016

  3. ICML 2020 (12th July to 18th July, )

    4. ICML 2020 offered a wealth of talks, workshops and tutorials. I focused in particular on the Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models workshop (), which included an interesting talk by Kyle Cranmer on how deep learning techniques were being applied in science. I also followed the excellent tutorial on Bayesian Deep Learning and a Probabilistic Perspective of Model Construction () by Andrew Wilson. The conference also offered mentoring sessions on career advice, emerging topics in machine learning, equality and more, which were particularly helpful for a PhD student.

  4. OxML 2020 (17th August to 25th August, )

    5. The school opened with two lectures closely related to my own research: Bayesian Machine Learning by Cheng Zhang and Gaussian Processes by James Hensman. These were followed by lectures on neural networks, natural language processing (NLP), computer vision, representation learning, causal machine learning, reinforcement learning and more. Alongside the lectures, there were tutorials and unconference sessions in which participants took the initiative to discuss specific topics in depth.

  5. GPSS 2020 (14th September to 17th September, )

    6. The school featured lectures on many aspects of Gaussian Processes (GPs), such as scalable GPs and deep GPs, as well as related topics including Bayesian optimisation, kernel design, Bayesian neural networks and composite GPs. Carl Henrik Ek gave a clear and engaging introduction to GPs.

All posts