Skip navigation
The Australian National University
photo of Peter

Dr Peter SUNEHAG

Position:Research Fellow
Email:JavaScript must be enabled to display this email address.
Phone:+61 2 612 57709
Building:RSISE (115)
Room:B224
Groups:CS, AI, NICTA
Staff category:Academic

Research interests:

I am working on theoretical foundations for machine learning and AI, in particular for reinforcement learning. Below follows some of my peer-reviewed publications:

A dual process theory of optimistic cognition, P. Sunehag, M. Hutter, CogSci'2014

Intelligence as Inference, P. Sunehag, M. Hutter, AGI'2014

Q-Learning for history based reinforcement learning, M. Daswani, P. Sunehag, M. Hutter ACML'2013

Conc. and Conf. for Discrete Bayesian Seq. Predictors, T. Lattimore, M. Hutter, P. Sunehag ALT'2013

The Sample-Complexity of General RL, T. Lattimore, M. Hutter, P. Sunehag ICML'2013

Feature Selection for Brain Computer Interfaces , G. Oliver,P. Sunehag,T. Gedeon CCMB'2013

Reinforcement Learning Agent with Evolving Hypothesis Classes, P. Sunehag, M. Hutter, AGI'2013

Rules of induction and the raven paradox in Bayesian confirmation theory, H. Afshar, P. Sunehag MaxEnt'2013

On ensemble methods for AIXI approximation, J. Veness, P. Sunehag, M. Hutter, AGI'2012

Optimistic Agents are Asymptotically Optimal, P. Sunehag, M. Hutter AusAI'2012

Coding of Non-Stationary Sources ..., P. Sunehag, W. Shao, M. Hutter AusDM'2012

Feature RL using Looping Suffix Trees, M. Daswani, P. Sunehag, M. Hutter EWRL'2012

Context Tree Maximizing Reinforcement Learning, P. Nguyen, P. Sunehag, M. Hutter, AAAI'2012

Asynchronous BCI Using Hidden Semi-Markov Models, G. Oliver, P. Sunehag, T. Gedeon, EMBC'2012

Recursive Channel Selection Techniques for BCI, G. Oliver, P. Sunehag, T. Gedeon, EMBC'2012

Adaptive Context Tree Weighting, A. O'Neill, M. Hutter, W. Shao, P. Sunehag, DCC'2012

Principles of Solomonoff Induction and AIXI, P. Sunehag, M. Hutter Solomonoff Memorial 2011

(Non)Equivalence of Universal Priors, I. Wood.,P. Sunehag, M. Hutter Solomonoff Memorial 2011

Axioms for Rational Reinforcement Learning, P. Sunehag, M. Hutter ALT'2011

Loss functions for improved on-policy control, M. Robards, P. Sunehag, Proc. of EWRL'2011

Feature Reinforcement Learning in Practice, P. Nguyen, P. Sunehag, M. Hutter, Proc. of EWRL'2011

Sparse Kernel-SARSA with an Elig. Trace, M. Robards, P. Sunehag, S. Sanner, B. Marthi ECML'2011

Consistency of Feature Markov Processes P. Sunehag, M. Hutter ALT'2010

Wearable-sensor activity analysis using semi-Markov models with a grammar O. Thomas, P. Sunehag, G. Dror et al. Pervasive and Mobile Computing 6(3): 342-350, 2010

Semi-markov kmeans clustering and activity recognition from ... M. Robards and P. Sunehag ICDM'2009

Variable Metric Stoch. Approx. Th. P. Sunehag et. al. AISTATS'2009

Real method of interp. on subcouples ... (with S. Astashkin) Studia Math. 185 (2008), 151-168

Using two-stage word frequency models ... P. Sunehag AISTATS'2007

Induced Graph Semantics: Another ... Hammersley-Clifford Thm T. Sears and P. Sunehag MaxEnt'2007

The real interp. method on couples of inters. (with S. Astashkin) Func. An. and Appl. 40 (2006) 218-221

... Interpolation of operators that almost agree P. Sunehag Journal of Approx. Th. 130 (2004) 78-98

Interpolation Of Banach Algebras And Tensor Products Of Banach Couples (with S. Kaijser) Journal of Mathematical Analysis and Applications 278 (2003) 367-375

Interpolation of Banach algebras and the unit problem. (with S. Kaijser) In: Function Spaces, Interpolation Theory and Related Topics, de Gruyter, Berlin (2002) 345-354.

Bio:

I did my Ph.D. in theoretical mathematics at Uppsala University in Sweden. The topic of the thesis was interpolation of Banach spaces and Banach Algebras. I then worked at NICTA in Canberra Australia as a machine learning researcher. I worked with Document Analysis, Optimization and activity recognition from body-worn sensor. I joined the ANU as a research fellow in July 2009 where I am primarily working on reinforcement learning, particularly on generic reinforcement learning.

Research opportunities with Dr Peter SUNEHAG

Responsible Officer:  JavaScript must be enabled to display this email address. / Page Contact:  JavaScript must be enabled to display this email address.