Optimal Sparse Recovery & Learning Latent Dynamics Models for Planning

Please note that Photo ID will be required. Please can attendees ensure their meetup profile name includes their full name to ensure entry.

Agenda:

• 18:30: Doors open, pizza, beer, networking
• 19:00: First talk
• 19:45: Break & networking
• 20:00: Second talk
• 20:45: Close

Sponsors
Man AHL: At Man AHL, we mix machine learning, computer science and engineering with terabytes of data to invest billions of dollars every day.

Evolution AI: Machines that Read - get answers from your text data.

• Title: Implicit Regularization for Optimal Sparse Recovery (Patrick Rebeschini - University of Oxford)

Abstract: Ridge regression is a fundamental paradigm in machine learning and statistics, and it has long been known to be closely connected to the implicit regularization properties of gradient descent methods, cf. early stopping. This connection has sparked research into a variety of directions aimed at developing computationally efficiency estimators, including acceleration, mini-batching, averaging, sketching, sub-sampling, preconditioning, and decentralization. Sparse recovery is another cornerstone of modern statistics and learning frameworks. Yet, perhaps surprisingly, here the connection to implicit regularization is not as well developed. Most results in the literature only involve limit statements (holding at convergence, for infinitesimal step sizes), apply to regimes with no (or limited) noise, and do not focus on computational efficiency. We address the following question: Can we establish an implicit regularization theory for gradient descent to yield optimal sparse recovery in noisy settings, achieving minimax rates with the same cost of reading the data? We will highlight the key ideas to obtain some first results here, along with a few related research directions.
(Based on joint work with Tomas Vaškevičius and Varun Kanade)

Bio: Patrick Rebeschini is an Associate Professor in the Department of Statistics at the University of Oxford. Patrick holds a PhD in Ops Research and Financial Engineering from Princeton University. Before joining Oxford, Patrick was at Yale University as an Assoc. Research Scientist in the Electrical Engineering Department, a Lecturer in the Computer Science Dept, and a Postdoctoral Associate in the Yale Institute for Network Science. His research is about discovering & exploiting fundamental principles in high-dimensional probability, statistics & optimization to design computationally efficient & statistically optimal algorithms for machine learning & artificial intelligence.

• Title: Learning Latent Dynamics Models for Planning (Timothy Lillicrap Google DeepMind)

• Abstract: Recent progress in deep reinforcement learning has led to extraordinary solutions to difficult control problems. Some of the best known, such as playing human-level Atari and winning at the games of Dota and Starcraft, have been achieved via model-free approaches. These approaches train value functions to anticipate future reward without explicitly representing potential action sequences. On the other hand, results in domains such as chess, Go, and shogi use model-based approaches that take advantage of a perfect model of the game to plan actions and estimate future reward. Here, I will introduce a new algorithm for planning with learned latent models and describe compelling results on continuous control domains that are controlled via high-dimensional pixel observations.

• Bio: Timothy Lillicrap received an Hon. B.Sc. in Cognitive Science & Artificial Intelligence from the University of Toronto and a PhD in Systems Neuroscience from Queen’s University in Canada. He moved to the University of Oxford in 2012 where he worked as a Postdoctoral Research Fellow. In 2014 he joined Google DeepMind as a Research Scientist. His research focuses on machine learning for optimal control and decision making, as well as using these mathematical frameworks to understand how the brain learns.