Mark Rowland

Research Scientist | DeepMind

About Me

I'm a Research Scientist at DeepMind. I'm interested in many areas across statistics, probability, reinforcement learning and optimisation, and interactions with areas of pure maths such as group theory and optimal transport theory.

Prior to joining DeepMind, I was a PhD student at Clare College, Cambridge University, co-supervised by Rich Turner at the Machine Learning Group and John Aston at Statslab, where I worked primarily on inference algorithms for discrete graphical models and Monte Carlo methods for kernel machines. Before my PhD, I studied for a BA and MMath at Cambridge, and wrote my Part III essay on random walks over the symmetric group, under the supervision of Nathanaël Berestycki.

News

(June 2020) Two papers accepted to ICML 2020 and a new preprint on arXiv: Revisiting Fundamentals of Experience Replay, Fast Computation of Nash Equilibria in Imperfect Information Games, and The Value-Improvement Path: Towards Better Representations for Reinforcement Learning.

Publications and Preprints

The Value-Improvement Path: Towards Better Representations for Reinforcement Learning

Will Dabney, André Barreto, Mark Rowland, Robert Dadashi, John Quan, Marc G. Bellemare, David Silver

[preprint]

Revisiting Fundamental of Experience Replay

William Fedus*, Prajit Ramachandran*, Rishabh Agarwal, Yoshua Bengio, Hugo Larochelle, Mark Rowland, Will Dabney [*equal contribution]

ICML 2020

[paper]

Fast Computation of Nash Equilibria in Imperfect Information Games

Rémi Munos, Julien Pérolat, Jean-Baptiste Lespiau, Mark Rowland, Bart De Vylder, Marc Lanctot, Finbarr Timbers, Daniel Hennes, Shayegan Omidshafiei, Audrunas Gruslys , Mohammad Gheshlaghi Azar, Edward Lockhart, Karl Tuyls

ICML 2020

[paper]

Navigating the Landscape of Games

Shayegan Omidshafiei*, Karl Tuyls*, Wojciech M. Czarnecki, Francisco C. Santos, Mark Rowland, Jerome Connor, Daniel Hennes, Paul Muller, Julien Pérolat, Bart De Vylder, Audrunas Gruslys, Rémi Munos [*equal contribution]

[preprint]

From Poincaré Recurrence to Convergence in Imperfect Information Games: Finding Equilibrium via Regularization

Julien Pérolat, Rémi Munos, Jean-Baptiste Lespiau, Shayegan Omidshafiei, Mark Rowland, Pedro Ortega, Neil Burch, Thomas Anthony, David Balduzzi, Bart De Vylder, Georgios Piliouras, Marc Lanctot, Karl Tuyls

[preprint]

Adaptive Trade-Offs in Off-Policy Learning

Mark Rowland*, Will Dabney*, Rémi Munos [*equal contribution]

To appear at AISTATS 2020

[preprint]

Conditional Importance Sampling for Off-Policy Learning

Mark Rowland, Anna Harutyunyan, Hado van Hasselt, Diana Borsa, Tom Schaul, Rémi Munos, Will Dabney

To appear at AISTATS 2020

[preprint]

A Generalized Training Approach for Multiagent Learning

Paul Muller, Shayegan Omidshafiei, Mark Rowland, Karl Tuyls, Julien Pérolat, Siqi Liu, Daniel Hennes, Luke Marris, Marc Lanctot, Edward Hughes, Zhe Wang, Guy Lever, Nicolas Heess, Thore Graepel, Rémi Munos

ICLR 2020

[paper]

Multiagent Evaluation under Incomplete Information

Mark Rowland*, Shayegan Omidshafiei*, Karl Tuyls, Julien Pérolat, Michal Valko, Georgios Piliouras, Rémi Munos [*equal contribution]

NeurIPS 2019

[paper]

α-Rank: Multi-Agent Evaluation by Evolution

Shayegan Omidshafiei*, Christos Papadimitriou*, Georgios Piliouras*, Karl Tuyls*, Mark Rowland, Jean-Baptiste Lespiau, Wojciech M. Czarnecki, Marc Lanctot, Julien Pérolat, Rémi Munos [*equal contribution]

Scientific Reports

[paper]

Meta-learning of Sequential Strategies

Pedro A. Ortega, Jane X. Wang, Mark Rowland, Tim Genewein, Zeb Kurth-Nelson, Razvan Pascanu, Nicolas Heess, Joel Veness, Alex Pritzel, Pablo Sprechmann, Siddhant M. Jayakumar, Tom McGrath, Kevin Miller, Mohammad Azar, Ian Osband, Neil Rabinowitz, András György, Silvia Chiappa, Simon Osindero, Yee Whye Teh, Hado van Hasselt, Nando de Freitas, Matthew Botvinick, Shane Legg

DeepMind Tech Report

[paper]

Statistics and Samples in Distributional Reinforcement Learning

Mark Rowland, Robert Dadashi, Saurabh Kumar, Rémi Munos, Marc G. Bellemare, Will Dabney

ICML 2019

[paper]

Unifying Orthogonal Monte Carlo Methods

Krzysztof Choromanski*, Mark Rowland*, Wenyu Chen, Adrian Weller [*equal contribution]

ICML 2019

[paper]

Antithetic and Monte Carlo Kernel Estimators for Partial Rankings

Maria Lomeli, Mark Rowland, Arthur Gretton, Zoubin Ghahramani

Statistics and Computing, 2019

[paper]

Orthogonal Estimation of Wasserstein Distances

Mark Rowland*, Jiri Hron*, Yunhao Tang*, Krzysztof Choromanski, Tamas Sarlos and Adrian Weller [*equal contribution]

AISTATS 2019

[paper]

Gaussian Process Behaviour in Wide Deep Neural Networks

Alexander G. de G. Matthews, Mark Rowland, Jiri Hron, Richard E. Turner, Zoubin Ghahramani

Extended version of the paper appearing at ICLR 2018, in submission

[preprint]

Geometrically Coupled Monte Carlo Sampling

Mark Rowland*, Krzysztof Choromanski*, François Chalus, Aldo Pacchiano, Tamas Sarlos, Richard E. Turner, Adrian Weller [*equal contribution]

NeurIPS 2018

[paper]

Structured Evolution with Compact Architectures for Scalable Policy Optimization

Krzysztof Choromanski*, Mark Rowland*, Vikas Sindhwani, Richard E. Turner, Adrian Weller [*equal contribution]

ICML 2018

[paper] [poster]

Gaussian Process Behaviour in Wide Deep Neural Networks

Alexander G. de G. Matthews, Jiri Hron, Mark Rowland, Richard E. Turner, Zoubin Ghahramani

ICLR 2018

[paper] [see above for journal version]

An Analysis of Categorical Distributional Reinforcement Learning

Mark Rowland, Marc G. Bellemare, Will Dabney, Rémi Munos, Yee Whye Teh

AISTATS 2018

[paper]

The Geometry of Random Features

Krzysztof Choromanski*, Mark Rowland*, Tamas Sarlos, Vikas Sindhwani, Richard E. Turner, Adrian Weller [*equal contribution]

AISTATS 2018

[paper]

Distributional Reinforcement Learning with Quantile Regression

Will Dabney, Mark Rowland, Marc G. Bellemare, Rémi Munos

AAAI 2018

[paper]

Uprooting and Rerooting Higher-Order Graphical Models

Mark Rowland*, Adrian Weller* [*equal contribution]

NIPS 2017

[paper] [poster]

The Unreasonable Effectiveness of Structured Random Orthogonal Embeddings

Krzysztof Choromanski*, Mark Rowland*, Adrian Weller [*equal contribution]

NIPS 2017

[paper] [poster]

Magnetic Hamiltonian Monte Carlo

Nilesh Tripuraneni, Mark Rowland, Zoubin Ghahramani, Richard E. Turner

ICML 2017

[paper] [supplementary material] [poster]

Conditions Beyond Treewidth for Tightness of Higher-Order LP Relaxations

Mark Rowland, Aldo Pacchiano, Adrian Weller

AISTATS 2017

[paper] [supplementary material] [poster]

Also presented at the International Conference on Principles and Practices of Constraint Programming (CP 2017)

Black-box Alpha Divergence Minimization

Yingzhen Li, José Miguel Hernández-Lobato, Mark Rowland, Daniel Hernández-Lobato, Thang Bui, Richard E. Turner

ICML 2016

[paper] [supplementary material]

Tightness of LP Relaxations for Almost Balanced Models

Adrian Weller, Mark Rowland, David Sontag

AISTATS 2016

[paper] [supplementary material] [poster]

Also presented at the International Conference on Principles and Practices of Constraint Programming (CP 2016)

Teaching

During my PhD, I gave supervisions on Part IA Groups, Part IA Probability, Part IB Linear Algebra, and Part IB Statistics. I also gave examples classes for Part III Bayesian Modelling and Computation in Lent 2017, and ran MATLAB introduction sessions for Maths Tripos students.

Contact

firstnamelastname [at] google.com