i++ School Newsletter

Week commencing 11 July 2011

Previous Newsletters

gamelab 2011

Professor Richard Bartle recently spoke at the Gamelab 2011 event in Barcelona.  In his keynote talk, he explained why social games were barely either social or games, and outlined why the people who are playing them today won't be playing them five years from now.  He did all this in a hall so noisy that even English-speakers wore translation headsets so they could hear what he was saying.  The slides from the talk are available here: http://www.mud.co.uk/richard/Barcelona.pdf

ccfea msc presentations

CCFEA recently held their annual MSc presentations.  This year's event was well received, and the overall standard of presentations was very high.  Four students in particular were awarded prizes for their outstanding presentation skills:
Lasse Pagh Frandsen ("Robust Portfolio Optimization")
Daniel Schiermer ("A Framework for Back-testing Algorithmic Trading Strategies Within the Marketcetera Platform")
Jose Javier Vargas Macias ("New Ways to Forecast Using Directional Changes with Dynamic Threshold and Evolutionary Computing")
Waed Zineddin ("Is Pairs Trading affected by Business Cycles?")

Each winner will receive a certificate and a cheque for 50 pounds. 

exciting openings with halcyon molecular in machine learning

Halcyon Molecular are working on the problem of sequencing from the Electron Microscope images of labeled DNA.
They need someone who would like to work with them in Silicon Valley, to develop highly accurate, fast, cheap DNA sequencing with the aim to apply the results of this development to a series of projects that aim at defeating all disease and death. They need:

• Someone with plenty of machine learning experience, who enjoys working with Bayesian models such as Hidden Markov Models. This person would be able to come up with new approaches to the sequencing challenge that could challenge the thinking of resident experts.
• A somewhat more junior machine learning specialist, but with similar knowledge. This person would be able to help explore a multiplicity of nuances of the models and work in close collaboration with everyone in the analysis team.
  
• A person with good image processing/analysis experience. Ideally, this person should also be familiar with GPU programming.
  
• Someone who is very good at building computing clusters, who knows enough programming to be able to set up a general image processing pipeline to best utilize the cluster, etc.
  
• Someone who would enjoy learning to collect and work with EM images. This should be someone with sufficient related expertise, but not necessarily an EM expert (though it can be).
  

If interested please contact Randal Koene at r@halcyonmolecular.com

PAPER PUBLISHED

Alexei Vernitski & Artem Pyatkin, "Astral graphs (threshold graphs), scale-free graphs and related algorithmic questions"

Abstract: The astral index of a graph is defined as the smallest number of astral graphs (also known as threshold graphs) into which the graph can be decomposed, divided by the number of vertices in the graph. The astral index is a promising new graph measure for analysing the structure of graphs in applications. In this paper, we prove some theoretical results concerning astral graphs and the astral index. We reveal a connection between astral graphs and scale-free graphs. We prove that finding the exact value of the astral index is an NP-complete problem.

PAPER PUBLISHED

Gower, E. and Hawksford, M.O.J., “Learning Overcomplete Dictionaries using a Cauchy Mixture Model for Sparse Decay”, World Academy of Science Engineering and Technology (WASET), International Journal of Electrical and Electronic Engineering, vol. 5, no. 2, March 2011, pp 88-95

Abstract: An algorithm for learning an overcomplete dictionary using a Cauchy mixture model for sparse decomposition of an underdetermined mixing system is introduced. The mixture density function is derived from a ratio sample of the observed mixture signals where 1) there are at least two but not necessarily more mixture signals observed, 2) the source signals are statistically independent and 3) the sources are sparse. The basis vectors of the dictionary are learned via the optimization of the location parameters of the Cauchy mixture components, which is shown to be more accurate and robust than the conventional data mining methods usually employed for this task. Using a well known sparse decomposition algorithm, we extract three speech signals from two mixtures based on the estimated dictionary. Further tests with additive Gaussian noise are used to demonstrate the proposed algorithm’s robustness to outliers.
 
For information Ephraim is now a lecturer in Botswana:
E. S. Gower is with the Department of Electrical Engineering, Faculty of Engineering, University of Botswana, Gaborone, Botswana, email:ephraim.gower@mopipi.ub.bw