[Humans of FDL] Tim Rudner
Meet EO Researcher Tim Rudner
Tim explains reinforcement learning, what he loves about the European Space Agency, and how Star Wars brought him to FDL Europe.
FDL: Can you tell us a bit about your background and what you’re working on?
TR: I'm originally from Cologne, Germany, but I’ve been living abroad (mainly in the United States and the UK) for the past six years. I studied mathematics and economics at Yale University and statistics at the University of Oxford, where I’m currently pursuing a PhD in Computer Science with a focus on probabilistic machine learning. My PhD research spans Bayesian deep learning and reinforcement learning, an area of machine learning concerned with finding algorithms that enable artificial agents to act optimally in some physical or virtual environment. In layman's terms, the core idea of reinforcement learning is to enable artificial agents to learn from feedback. For example, if you want to train your dog to perform a trick, you would reward it with a treat every time it performs the trick correctly. By giving your dog a reward every time it ‘succeeds’ during training, it learns to distinguish between ‘good’ and ‘bad’ actions (shaking paws vs. not shaking paws) given certain conditions and states. Reinforcement learning is based on the same principle: reinforcement learning researchers try to find algorithms that allow artificial agents to learn best actions from a continuous stream of feedback in complex environments.
FDL: So as an AI researcher, how did you enjoy diving into the world of earth observation?
TR: My time with FDL has been an amazing learning experience. We started by spending a week at the European Space Agency’s (ESA) Phi-Lab just outside of Rome, where we engaged with researchers from academia and industry, who use satellite imagery for a wide range of application areas. During that week, I learned so much about the challenges and opportunities associated with earth observation (EO) data. I learned about core concepts in EO and EO terminology, something that should should allow me to effectively apply machine learning methods to any type of earth observation data while avoiding common pitfalls.
I loved working at ESA, and I thought everyone at ESA’s Phi-Lab was incredibly helpful and knowledgeable. During our stay there, we were fortunate to meet Josef Aschbacher, Director of ESA’s Earth Observation Programs, and to speak with him about the Agency’s vision for the future of its earth observation activities. I think it’s really exciting that ESA’s Copernicus Program aims to achieve a global, continuous, and autonomous Earth observation capacity, and I love that ESA is making all of this data freely available. There is a wealth of information and insights that could potentially improve thousands of people’s lives hidden in this data, and it’s just waiting to be discovered!
FDL: What’s your favorite movie?
TR: I’ve always been a big Star Wars fan! I’ve been fascinated by space from a young age, and this fascination for pretty much anything space just stuck with me over the years. After finishing high school, I even considered studying astrophysics, but ultimately decided to study mathematics instead, because I felt that it was more generally applicable. Since then, it’s been a dream of mine to work on space-related topics, and although FDL is not quite intergalactic space travel, working with satellite imagery to learn more about our planet has been a super exciting experience!
FDL What is your proudest accomplishment/achievement?
TR: I feel extremely grateful for the many opportunities I’ve been afforded over the course of my life and especially over the past six years. As a first-generation college student, being able to attend Yale University on a full scholarship and winning a Rhodes Scholarship to pursue graduate study at the University of Oxford felt--and still feels--like a dream, and I consider them my proudest achievements to date. However, while I feel very humbled and fortunate for having been able to attend Yale and to be selected for a Rhodes Scholarship, neither would have been possible without the love and endless support from my mother, who always believed in me and encouraged me to pursue my dreams. My achievements are as much mine as they are hers, and I feel incredibly lucky to have her as my role model.
About Tim Rudner
Tim G. J. Rudner is a PhD student in Computer Science at the University of Oxford, where he conducts research on reinforcement learning and Bayesian deep learning. He received a master’s degree in statistics from the University of Oxford and an undergraduate degree in mathematics and economics from Yale University. Prior to starting his PhD, Tim conducted research on game theoretic equilibria in digital goods markets, systemic stress in the financial sector, and drivers of financial crises. He has been involved in ‘AI for Good’ projects with the Rhodes Artificial Intelligence Lab and the Norwegian Refugee Council and worked as a research consultant at the European Central Bank. Tim strives to use machine learning to understand and model the emergence of cooperation in complex multi-agent systems, to create machine learning tools that benefits everyone, and to help advance policy discussions on machine intelligence.