Marc Deisenroth's Homepage

	Department of Engineering Computational and Biological Learning Lab

University of Cambridge > Department of Engineering > Information Engineering > Computational and Biological Learning Lab > Marc Deisenroth

Home

Publications

Downloads

Talks

Vita

Marc Deisenroth's homepage

	Visiting Ph.D. Student Department of Engineering University of Cambridge, UK
	Ph.D. Candidate Faculty of Informatics Universität Karlsruhe (TH), Germany

Research Interests

I'm especially interested in

Bayesian inference and learning
reinforcement learning
Gaussian process models
optimal control
active learning/optimal design
inference and learning in nonlinear dynamical systems
signal processing

News

Workshop paper accepted (2009-09-02)

Our paper on Efficient Reinforcement Learning for Motor Control got accepted.

NIPS workshop accepted (2009-08-26)

Our NIPS workshop on Probabilistic Approaches for Robotics and Control got accpeted.

ICML paper accepted and online (2009-04-22)

The paper is about Kalman filtering in nonlinear dynamic systems using Gaussian process models. More information and the paper itself are here.

Gaussian Process Dynamic Programming published (2009-03-15)

The original article (requires access to ScienceDirect) is here. A pre-print is freely available online.

ALGPDP code online (2009-01-31)

A preliminary version of the ALGPDP code is online.

Videos of Double Pendulum Controller (2009-01-12)

I put up flash-videos of the learning process.

Double Pendulum Controller (2008-12-16)

Using probabilistic inference, we can learn to swing up and balance a double pendulum attached to a cart (from scratch). Due to appropriate handling of uncertainties in simulation, we only need few interactions with the real system.

Hardware Experiment (2008-11-25)

Our probabilistic algorithm for fast learning also works in practice with a cart-pole swing-up task. Here is a very brief introduction, what's going on, including snapshots of a hardware experiment and the link to the corresponding video.

Probabilistic RL Algorithm (2008-07-15)

We recently developed a probabilistic reinforcement learning algorithm. An intuitive description and a bunch of videos are here. A paper describing the ideas will be published soon in Lecture Notes on Computer Science. An electronic copy is available here.

Contact Information

Department of Engineering
Trumpington Street
Cambridge CB2 1PZ, UK
Room BE4-36
Phone: +44 (0) 122 374 851 2
Fax: +44 (0) 1 223 332 662
Email: mpd37 [AT] cam.ac.uk

MPI for Biological Cybernetics, Tübingen, Germany | ISAS, Universität Karlsruhe (TH), Germany