The 2013 Classic Paper Prize at the International Conference on Machine Learning (ICML) was won by Zoubin Ghahramani and coauthors Xiaojin Zhu and John Lafferty for their 2003 paper “Semi-Supervised Learning Using Gaussian Fields and Harmonic Functions” . The Classic Paper Prize is given to the paper  published at ICML 10 years previously which has had the most impact on the field. Semi-supervised learning refers to the problem of combining small amounts of labelled data (i.e. supervised learning) with large amounts of unlabelled data (i.e. unsupervised learning).
This 2003 paper, which has now been cited over 1400 times, developed a simple and highly-scalable graph-based method for semi-supervised classification, and related it to harmonic functions, random walks, electric networks, and spectral graph theory. Graph-based semi-supervised learning has now become a standard approach for combining labelled and unlabelled data in many application domains.

Link to the paper: http://mlg.eng.cam.ac.uk/zoubin/papers/zgl.pdf
Link to the ICML conference: http://icml.cc/2013/
Link to the Machine Learning group website: http://mlg.eng.cam.ac.uk/

Group post-doc Christian Steinruecken radio interviews with The Naked Scientists, an award-winning BBC weekly radio programme delivered by a University-based group focusing on broad topics in science for a general audience.  Christian spoke to The Naked Scientists about data compression, some basics of how it works, and its role in the technologies that we employ today.

Learn more about The Naked Scientists and their programme on their webpage.

Information about Christian and his research can be found on his webpage.

The department press release with some excerpts from the interview here.

The original interview can be found here.

You can listen to the interview as an mp3 here.

Eight new papers from the group are to appear in the proceedings of the 30th International Conference on Machine Learning (ICML 2013), to be held in Altanta, Georgia, USA in June.  ICML is a leading conference on machine learning.  Here are the list of papers with links to the documents:

• D Duvenaud, JR Lloyd, R Grosse, JB Tenenbaum, and Z Ghahramani.  Structure Discovery in Nonparametric Regression through Compositional Kernel Search.  [arXiv]
• E Gilboa, Y Saatci, and JP Cunningham.  Scaling multidimensional Gaussian Processes using projected additive approximations.  [arXiv]
• C Heaukulani, and Z Ghahramani.  Dynamic Probabilistic Models for Latent Feature Propagation in Social Networks.  [pdf]
• B Lakshminarayanan, DM Roy, YW Teh.  Top-down particle filtering for Bayesian decision trees.  [arXiv]
• D Lopez-Paz, JM Hernandez-Lobato, and Z Ghahramani.  Gaussian process vine copulas for multivariate dependence.  [pdf]
• C Reed and Z Ghahramani.  Scaling the Indian Buffet Process via Submodular Maximization.  [arXiv]
• AG Wilson and RP Adams.  Gaussian Process Covariance Kernels for Pattern Discovery and Extrapolation.  [arXiv]
• Y Wu, JM Hernandez-Lobato, and Z Ghahramani.  Dynamic Covariance Models for Multivariate Financial Time Series.

Abstracts and additional material can be found on our publications page, and links to author webpages can be found on our group members page.

Congratulations to Carl Rasmussen for being awarded the 4th Year Best Lecturer Award in the Engineering department for his teaching in the course “4F13: Machine Learning”.  You can find details about the course on our homepage.  The excellence in teaching award is voted on by 4th year students in the department.

Richard Turner has been appointed University Lectureship in Computer Vision and Machine Learning. His research programme spans computer perception, signal processing, machine learning, and neuroscience. He’ll strengthen connections between the Computational and Biological Learning Lab, the Machine Intelligence Lab and the Signal Processing Lab.

Richard Turner’s website: http://www.gatsby.ucl.ac.uk/~turner/

Prof. John Cunningham’s article “Neural population dynamics during reaching”, work completed while a Post-Doc in the Machine Learning Group, is published in the journal Nature. From the abstract:

“Most theories of motor cortex have assumed that neural activity represents movement parameters. This view derives from what is known about primary visual cortex, where neural activity represents patterns of light. Yet it is unclear how well the analogy between motor and visual cortex holds. Single-neuron responses in motor cortex are complex, and there is marked disagreement regarding which movement parameters are represented. A better analogy might be with other motor systems, where a common principle is rhythmic neural activity. Here we find that motor cortex responses during reaching contain a brief but strong oscillatory component, something quite unexpected for a non-periodic behaviour. Oscillation amplitude and phase followed naturally from the preparatory state, suggesting a mechanistic role for preparatory neural activity. These results demonstrate an unexpected yet surprisingly simple structure in the population response. This underlying structure explains many of the confusing features of individual neural responses.”

The article: http://www.nature.com/nature/journal/vaop/ncurrent/full/nature11129.html

Press articles: Department; The Atlantic.

Prof. John Cunningham’s website at Washington University in St. Louis: http://cunningham.wustl.edu/

PhD student Yarin Gal wins prestigious Google Europe Doctoral Fellowship in Machine Learning. This year 15 fellowships were awarded to new doctoral students across Europe. From the ‘Research at Google’ website:

“We introduced the Google Fellowship programme in 2009 in the United States to broaden our support of university research. Universities continue to be the source of some of the most innovative research in computer science, and in particular it is the students that they foster who are the future of our field. In 2010, we extended the fellowship programme to Universities in Europe.”

Press release: http://research.google.com/university/relations/fellowship_recipients.html

PhD student James Lloyd places third out of over 100 competitors in the Load Forecasting track of the Global Energy Forecasting Competition 2012 (GEFCom2012), sponsored by the IEEE Power & Energy Society. From the competition website:

“This competition will bring together the state-of-the-art techniques for energy forecasting, serve as the bridge to connect academic research and industry practice, promote analytics in power engineering education, and prepare the industry to overcome forecasting challenges in the smart grid world.”

Competition website and leaderboard: http://www.kaggle.com/c/global-energy-forecasting-competition-2012-load-forecasting

James Lloyd’s website: http://mlg.eng.cam.ac.uk/lloyd/

Dan Roy‘s PhD dissertation, “Computability, inference and modeling in probabilistic programming”, has been selected as a winner of a MIT/EECS Sprowls Award, recognizing it as one of the best dissertations in Computer Science at MIT in 2011.

Simon Lacoste-Julien has been awarded a prestigious Research in Paris grant, and will take up further post-doctoral studies with Francis Bach in the SIERRA project team located in the joint INRIA/CNRS/Ecole Normale Superieure computer science laboratory, starting on September 1st 2011.

Read more →