Richard E. Turner PhD Student, Gatsby Computational Neuroscience Unit Alexandra House 17 Queen Square UCL London turner-at-gatsby.ucl.ac.uk

Introduction:
My work lies at the interface between neuroscience (which tries to understand the brain) and machine-learning (which tries to build algorithms which learn from data). The goal is to develop neuroscience inspired algorithms to solve problems which the brain solves routinely e.g. figuring out how many sound sources there are in an acoustic scene and what the individual contributions from each source are. The behaviour of these algorithms can then be compared to natural processing in the brain (e.g. neural activities or psychophysics) in order to better understand what they are doing.

Current Highlights:

• A draft version of my PhD Thesis, Statistical Models for Natural Sounds is now ready. There are also a number of related audio demonstrations.
• Awarded an EPSRC Postdoctoral Fellowship Grant at the Interface for the Life Sciences. This is a prestigious three year grant that I will spend at the CBL Lab in Cambridge with Zoubin Ghahramani and at the Lab for Computer Vision, NYU with Eero Simoncelli.
• A book chapter; Two problems with variational Expectation Maximisation for time-series models.
• Probabilistic Auditory Scene Analysis talk Presented at the Gordon Research Conference Sensory coding and the natural environment.
• There is a folk-theorem which says, "Variational approximations are always more compact than the true distribution". In this note we provide counter examples which disprove this folk theorem.
• On sparsity and overcompleteness in image models: A public talk presented at the Inference group, Cambridge. This work is based on our NIPS 2008 paper, On Sparsity and Overcompleteness in Image Models.
• Probabilistic Amplitude Demodulation, paper accepted and given an oral presentation at ICA 2007 conference, London. Winner of best student paper award.
• Advances in Models for Acoustic Processing NIPS workshop, Vancouver talk and paper: 'Modeling Natural Sounds with Gaussian Modulation Cascade Processes', R. Turner and M. Sahani 2006
• Accepted by Neural Computation: 'A maximum-likelihood interpretation for slow feature analyis', R. Turner and M. Sahani 2006

Workshops organised

• Advanced probabilistic techniques (APT) workshop A two hour lecture on a probabilistic method (loopy BP, EP, DBNs, DPs and SMCMS) followed by coding up the method on a toy problem in a pair
• Beyond Simple Cells (BSC) workshop Computational modelers of visual cortex meet the physiologists whose data they try and model.

Papers, talks and teaching

• Publications
• Public Talks
• Research Talks
• Tea Talks
• Journal Club Presentations
• Research Notes
• Teaching

Sounds

• Audio demonstrations

Things I used to organise...

• I have given up organising Tea now, but the current schedule is available here
• I used to run the Generative models, perception and natural statistics working group before Gatsby reverted to the old journal club system

Personal

• A collection of interesting problems
• Hyper-extension when skipping stones: Or why I am good at throwing things.
• Ultimate Photos