Short summary of my experience at ProbAI 2019

Brief intro

I attended a summer school about probabilistic machine learning in the first week of June this year in 2019. Probabilistic machine learning (sometimes Bayesian Machine Learning) is a discipline of machine learning where the uncertainty of a decision is also calculated. This is really interesting topic for many medical and other mission critical applied research now, as often a yes-or-no decision is not enough.

The venue

Trondheim is absoutely stunning, but not cheap. If you are looking for some untouched nature or a retreat from your PhD life, I definitely recommend it, but be wary - apart from nature there is not too much to see! However, all the restaurants were top notch in Trondheim. I would definitely recommend Peppes Pizza if you are into greasy Domino’s like Pizza.

A bird's eye view of Trondheim

Lectures I liked the most

The two talks that I definitely adored the most were the one about Importance Weighted Autoencoders and the one about Sparse Bayesian Neural Networks. I have done my Master’s thesis about automatic relevance determination, and I think it’s one of the most appealing way to do Bayesian for people, who are looking for a principled framework, dealing with prior probability distributions.

What I’ve learned

I think it was a great practical look at Bayesian Machine Learning and Probabilistic Graphical models. It was refreshing to have a great review of the past few years machine learning literature, and I got confirmed that cycle-consistent GANs (and GANS in general) are indeed becoming more than a niche ML research area.

The related coursework was one of the most rewarding assignments I have ever done, because it really helped me bring together the theory and practice in my mind.

Depending on next year’s topics I might consider attending again, but it was definitely a refreshing breakout from my PhD life.

As always, the best thing of summer schools is the ability to make new connections

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under Marie Sklodowska-Curie grant agreement No 766287.