How DID YOU GET INTO Machine LEARNING?
My career in ML was far from being a dream career choice I had since my early childhood. The last one involved me fighting the villains wearing a red cape somewhere in a small town in the middle of Greece. When I grew up enough to realise that red capes are something you mainly wear at costume parties, I decided to study Electrical Engineering just because most of the courses were physics and maths and those, I knew for sure, work all year around. I majored in Telecommunications and Networks but in my last year, I felt i needed a change. I decided to do my thesis on something completely different; Databases and Programming.
I was restless, I wanted to try new things, including living abroad. My supervisor Prof. Timos Sellis was the first one to suggest that I do a Masters degree abroad. Thus, I found myself leaving the South for the North, chasing a master in Informatics in Edinburgh. There I had my first ever course in Machine Learning. I found out that the Bayes rule was the approach to life I could relate to; deal with the uncertainty and learn from the evidence! After my master studies, I joined MSRC for an internship. That's where I came across Prof. Zoubin Ghahramani's lab working on the forefront of ML. I had never thought of a PhD before that. As a matter of fact, it was not until later that I found out what a PhD actually is. All I knew is that I liked research and the challenges. I was lucky, or better said, fortunate enough to end up in one of the most prestigious labs in ML and continue my postdoc research as part of Prof. Yee Whye's Teh lab in Oxford. All these wouldn't have been possible but for two main reasons; my constant eager for learning and trying new challenges and the great support of the amazing people I met along the way. I can see now that I never had my mind fixed on a choice. Rather, I was constantly changing, evolving. I had many second thoughts throughout the path, but always had people that supported and helped me; mentors, family, friends, animals but most of all, myself!
I was restless, I wanted to try new things, including living abroad. My supervisor Prof. Timos Sellis was the first one to suggest that I do a Masters degree abroad. Thus, I found myself leaving the South for the North, chasing a master in Informatics in Edinburgh. There I had my first ever course in Machine Learning. I found out that the Bayes rule was the approach to life I could relate to; deal with the uncertainty and learn from the evidence! After my master studies, I joined MSRC for an internship. That's where I came across Prof. Zoubin Ghahramani's lab working on the forefront of ML. I had never thought of a PhD before that. As a matter of fact, it was not until later that I found out what a PhD actually is. All I knew is that I liked research and the challenges. I was lucky, or better said, fortunate enough to end up in one of the most prestigious labs in ML and continue my postdoc research as part of Prof. Yee Whye's Teh lab in Oxford. All these wouldn't have been possible but for two main reasons; my constant eager for learning and trying new challenges and the great support of the amazing people I met along the way. I can see now that I never had my mind fixed on a choice. Rather, I was constantly changing, evolving. I had many second thoughts throughout the path, but always had people that supported and helped me; mentors, family, friends, animals but most of all, myself!
WhAT WILL YOU Be teaching?
This tutorial will be an overview of probabilistic reasoning with a focus on its connections to deep networks. We will show how fundamental principles of probabilistic reasoning are or can be found in deep neural networks, i.e. modelling, learning, inference, loss function, regularization etc. We will discuss that even though modern deep learning models used in practice do not capture model confidence they can easily extended to account for uncertainty through their extentions to probabilistic approaches. More specifically, we will show their close relation to a family of probabilistic models such as the Gaussian process. The overarching goal of the tutorial is for your to strengthen your ability to reason probabilistically using deep NNs.
What advice would you give to those getting started in machine/deep learning?
Don't be afraid to follow your interests in the field even if this includes a 180 degrees turn to the opposite direction every time. Don't be afraid to talk to senior people in the field. ML is a field that is constantly evolving and so we should too.