Brendan O'Donoghue

Brendan O'Donoghue, Ph.D.
Research Scientist at DeepMind
Former advisor: Professor Stephen Boyd
Google scholar profile

Contact: email, twitter, github

Education

Ph.D., M.S., Computer Science,
Stanford University, January 2013
B.A., M.A., M.Eng., Information and Computer Engineering,
Gonville and Caius College,
Cambridge University, June 2007

Interests and current research

Artificial intelligence
Machine learning
Reinforcement learning
Dynamic systems and control
Convex optimization

Software

SCS: Large-scale convex quadratic cone solver.

Journal Articles

POMRL: No-regret learning-to-plan with increasing horizons: K. Khetarpal, C. Vernade, B. O'Donoghue, S. Singh, and T. Zahavy
Transactions on Machine Learning Research (TMLR), 2023.

Operator splitting for a homogeneous embedding of the monotone linear complementarity problem: B. O'Donoghue
SIAM Journal on Optimization, 31(3), pp. 1999-2023, August 2021.

Globally convergent type-I Anderson acceleration for non-smooth fixed-point iterations: J. Zhang, B. O'Donoghue, and S. Boyd
SIAM Journal on Optimization, 30(4), pp. 3170-3197, November 2020.

Clinically applicable deep learning for diagnosis and referral in retinal disease: J. De Fauw, et al.
Nature Medicine, 24(9), pp. 1342-1350, August 2018.

Conic optimization via operator splitting and homogeneous self-dual embedding: B. O'Donoghue, E. Chu, N. Parikh, and S. Boyd
Journal of Optimization Theory and Applications, 169(3), pp. 1042-1068, June 2016.

Large-scale convex optimization for dense wireless cooperative networks: Y. Shi, J. Zhang, B. O'Donoghue, and K. Letaief
IEEE Transactions on Signal Processing, 63(18), pp. 4729-4743, September 2015.
IEEE 2016 SPS Young Author Best Paper Award.

Approximate dynamic programming via iterated Bellman Inequalities: Y. Wang, B. O'Donoghue, and S. Boyd
International Journal of Robust and Nonlinear Control, 25(10), pp. 1472-1496, July 2015.

Adaptive restart for accelerated gradient schemes: B. O'Donoghue and E. J. Candès
Foundations of computational mathematics, 15(3), pp. 715-732, June 2015.

Fast alternating direction optimization methods: T. Goldstein, B. O'Donoghue, S. Setzer, and R. Baraniuk
SIAM Journal on Imaging Sciences, 7(3), pp.1588-1623, August 2014.

A spread-return mean-reverting model for credit spread dynamics: B. O'Donoghue, M. Peacock, J. Lee, and L. Capriotti
International Journal of Theoretical and Applied Finance, 17(3), pp. 1-14, May 2014.

Performance bounds and suboptimal policies for multi-period investment: S. Boyd, M. Mueller, B. O'Donoghue, and Y. Wang
Foundations and Trends in Optimization, 1(1), pp. 1-69, January 2014.

A splitting method for optimal control: B. O'Donoghue, G. Stathopoulos, and S. Boyd
IEEE Transactions on Control Systems Technology, 21(6), pp. 2432-2442, November 2013.

Conference Articles

Probabilistic inference in reinforcement learning done right: J. Tarbouriech, T. Lattimore, and B. O'Donoghue
Advances in Neural Information Processing Systems (NeurIPS), 2023.

Optimistic meta-gradients: S. Flennerhag, T. Zahavy, B. O'Donoghue, H. van Hasselt, A. György, and S. Singh
Advances in Neural Information Processing Systems (NeurIPS), 2023.

Efficient exploration via epistemic-risk-seeking policy optimization: B. O'Donoghue
Proceedings of the International Conference on Machine Learning (ICML), 2023.

ReLOAD: Reinforcement learning with optimistic ascent-descent for last-iterate convergence in constrained MDPs: T. Moskovitz, B. O'Donoghue, V. Veeriah, S. Flennerhag, S. Singh, and T. Zahavy
Proceedings of the International Conference on Machine Learning (ICML), 2023.

The neural testbed: Evaluating joint predictions: I. Osband, Z. Wen, S. Asghari, V. Dwaracherla, B. Hao, M. Ibrahimi, D. Lawson, X. Lu, B. O'Donoghue, and B. Van Roy
Advances in Neural Information Processing Systems (NeurIPS), 2022.

Variational Bayesian optimistic sampling: B. O'Donoghue and T. Lattimore
Spotlight Advances in Neural Information Processing Systems (NeurIPS), 2021.

Reward is enough for convex MDPs: T. Zahavy, B. O'Donoghue, G. Desjardins, and S. Singh
Spotlight Advances in Neural Information Processing Systems (NeurIPS), 2021.

Variational Bayesian reinforcement learning with regret bounds: B. O'Donoghue
Advances in Neural Information Processing Systems (NeurIPS), 2021.

Practical large-scale linear programming using primal-dual hybrid gradient: D. Applegate, M. Díaz, O. Hinder, H. Lu, M. Lubin, B. O'Donoghue, and W. Schudy
Advances in Neural Information Processing Systems (NeurIPS), 2021.

Matrix games with bandit feedback: B. O'Donoghue, T. Lattimore, and I. Osband
Proceedings of the 37th Conference on Uncertainty in Artificial Intelligence, (UAI), 2021.

Discovering a set of policies for the worst case reward: T. Zahavy, A. Barreto, D. Mankowitz, S. Hou, B. O'Donoghue, I. Kemaev, and S. Singh
Spotlight Proceedings of the International Conference on Learning Representations (ICLR), 2021.

Sample Efficient Reinforcement Learning with REINFORCE: J. Zhang, J. Kim, B. O'Donoghue, and S. Boyd
Proceedings of the AAAI Conference on Artificial Intelligence, 35(12), 10887-10895, 2021.

Making sense of reinforcement learning and probabilistic inference: B. O'Donoghue, I Osband, and C. Ionescu
Spotlight Proceedings of the International Conference on Learning Representations (ICLR), 2020.

Hamiltonian descent for composite objectives: B. O'Donoghue and C. J. Maddison
Advances in Neural Information Processing Systems (NeurIPS), 2019.

Visualizations of decision regions in the presence of adversarial examples: G. Swirszcz, B. O'Donoghue, and P. Kohli
Debugging Machine Learning Models Workshop, ICLR, 2019.

Verification of non-linear specifications for neural networks: C. Qin, K. (Dj) Dvijotham, B. O'Donoghue, R. Bunel, R. Stanforth, S. Gowal, J. Uesato, G. Swirszcz, and P. Kohli
Proceedings of the International Conference on Learning Representations (ICLR), 2019.

Adversarial risk and the dangers of evaluating against weak attacks: J. Uesato, B. O'Donoghue, A. van den Oord, and P. Kohli
Proceedings of the International Conference on Machine Learning (ICML), pp. 5025-5034, 2018.

The uncertainty Bellman equation and exploration: B. O'Donoghue, I. Osband, R. Munos, and V. Mnih
Oral Proceedings of the International Conference on Machine Learning (ICML), pp. 3836-3845. 2018.

Combining policy gradient and Q-learning: B. O'Donoghue, R. Munos, K. Kavukcuoglu, and V. Mnih
Proceedings of the International Conference on Learning Representations (ICLR), 2017.

Iterated approximate value functions: B. O'Donoghue, Y. Wang, and S. Boyd
Proceedings European Control Conference, pp. 3882-3888, Zurich, July 2013.

Min-max approximate dynamic programming: B. O'Donoghue, Y. Wang, and S. Boyd
Proceedings IEEE Multi-Conference on Systems and Control, pp. 424-431, September 2011.

Other

On the connection between Bregman divergence and value in regularized Markov decision processes: B. O'Donoghue
Technical note, 2022.

Solving mixed integer programs using neural networks: V. Nair*, S. Bartunov*, F. Gimeno*, I. von Glehn*, P. Lichocki*, I. Lobov*, B. O'Donoghue*, N. Sonnerat*, C. Tjandraatmadja*, P. Wang*, et al.
(* Equal contribution). In submission, 2021.

Discovering diverse nearly optimal policies with successor features: T. Zahavy, B. O'Donoghue, A. Barreto, V. Mnih, S. Flennerhag, and S. Singh
Working draft, 2021.

Strength in numbers: Trading-off robustness and computation via adversarially-trained ensembles: E. Grefenstette, R. Stanforth, B. O'Donoghue, J. Uesato, G. Swirszcz, and P. Kohli
Working draft, 2018.

Hamiltonian descent methods: C. J. Maddison, D. Paulin, Y. W. Teh, B. O'Donoghue, and A. Doucet
Working draft, 2018.

Training verified learners with learned verifiers: K. (Dj) Dvijotham, S. Gowal, R. Stanforth, R. Arandjelovic, B. O'Donoghue, J. Uesato, and P. Kohli
Working draft, 2018.

A primal-dual operator splitting method for conic optimization: E. Chu, B. O'Donoghue, N. Parikh, and S. Boyd
Stanford internal report, (2013).

Ph.D. Thesis

Suboptimal control policies via convex optimization: B. O'Donoghue