The 2013 UK Space Design Competition Request for Proposals (RFP) is now out!
The UK Space Design Competition 2013 is open to all UK secondary school students in years 9-13. Teams must consist of between 8 and 12 students, plus a supervising adult, but need not be affiliated with any particular institution. This means that schools, colleges, science clubs, and societies are all free to enter a team, provided that the above criteria are satisfied. All team members must be specified in the initial application.
Check out http://uksdc.org for details!
Had a great time with the gang watching the Olympics closing ceremony. Coincidentally, we had another steak night. Yummy. Hats off to Kyu Hwa the Chef!
In other news, just submitted a paper on ARTY (and a trial-run by a child with special needs) to the IROS Workshop on robotic wheelchairs — fingers crossed that it’ll get accepted. Will put up a pre-print soon. If you are in the field, I recommend checking out the workshop on shared-control (organised by Tom Carlson at EPFL) at this year’s SMC.
Next week will be busy for us as we gear up for the Summer School.
Just submitted an IROS camera-ready copy of some recent work on online spatio-temporal learning:
In this work, we are primarily concerned with robotic systems that learn online and continuously from multi-variate data-streams. Our first contribution is a new recursive kernel, which we have integrated into a sparse Gaussian Process to yield the Spatio-Temporal Online Recursive Kernel Gaussian Process (STORK-GP). This algorithm iteratively learns from time-series, providing both predictions and uncertainty estimates. Experiments on benchmarks demonstrate that our method achieves high accuracies relative to state-of-the-art methods. Second, we contribute an online tactile classifier which uses an array of STORK-GP experts. In contrast to existing work, our classifier is capable of learning new objects as they are presented, improving itself over time. We show that our approach yields results comparable to highly-optimised offline classification methods. Moreover, we conducted experiments with human subjects in a similar online setting with true-label feedback and present the insights gained.
This work was nominated as a finalist for the 2012 CoTeSys Cognitive Robotics Best Paper Award.
After a couple of weeks of putting it off, I’m releasing an alpha version of the Online Temporal Learning (OTL) C++ library, which can be used for learning time-series in an online environment (samples are processed one at a time). The Online Echo State Gaussian Process (OESGP) and Spatial-Temporal Online Recursive Kernel Gaussian Process (STORK-GP) algorithms are part of this library.
Find the code at bitbucket: https://bitbucket.org/haroldsoh/otl
And be sure to check out the Getting Started Guides on the wiki.
Comments/Criticisms/Bug reports are welcome! Though, depending on my workload, replies may be slow in coming. Have fun!
P.S.: If you’re looking for the OESGP Experimental setup scripts, you can find them here.
I’ve released the MATLAB scripts used to generate the Multi-Reward POMDPs used in my papers. You can access them at the bitbucket repository:
The code is released under GPL. Comments/Criticisms/Bug reports are welcome. Have fun!
A little exhausted after a couple of paper submissions to IROS. Looking forward to a break for a few days while I re-organise and re-think. And some opportunity for reading (Kantz and Schreiber’s Nonlinear Time Series Analysis that I bought ages ago but haven’t finished, and Chaitin’s Meta Math that I have read but have largely forgotten).
Plus, my desk is clean again!
Paper accepted at IJCNN 2012; Never been to Brisbane!
Summary: In this work, we contribute the online echo state gaussian process (OESGP), a novel Bayesian-based online method that is capable of iteratively learning complex temporal dynamics and producing predictive distributions (instead of point predictions). Our method can be seen as a combination of the echo state network with a sparse approximation of Gaussian processes (GPs). Extensive experiments on the one-step prediction task on well-known benchmark problems show that OESGP produced statistically superior results to current online ESNs and state-of-the-art regression methods. In addition, we characterise the benefits (and drawbacks) associated with the considered online methods, specifically with regards to the trade-off between computational cost and accuracy. For a high-dimensional action recognition task, we demonstrate that OESGP produces high accuracies comparable to a recently published graphical model, while being fast enough for real-time interactive scenarios.