Past

1. Introduction and survey of the cohomological results

This will be a relatively gentle introduction to the topologist's point of view of Riemann's moduli space followed by a description of its rational and torsion cohomology for large genus.

 

 

/notices/events/abstracts/stochastic-analysis/mt06/bolley.shtml

 

 

Numerical weather prediction models require an estimate of the current state of the atmosphere as an initial condition. Observations only provide partial information, so they are usually combined with prior information, in a process called data assimilation. The dynamics of hazardous weather such as storms is very nonlinear, with only a short predictability timescale, thus it is important to use a nonlinear, probabilistic filtering method to provide the initial conditions. 

Unfortunately, the state space is very large (about 107 variables) so approximations have to be made.

The Ensemble Kalman filter (EnKF) is a quasi-linear filter that has recently been proposed in the meteorological and oceanographic literature to solve this problem. The filter uses a forecast ensemble (a Monte Carlo sample) to estimate the prior statistics. In this talk we will describe the EnKF framework and some of its strengths and weaknesses. In particular we will demonstrate a new result that not all filters of this type bear the desired relationship to the forecast ensemble: there can be a systematic bias in the analysis ensemble mean and consequently an accompanying shortfall in the spread of the analysis ensemble as expressed by the ensemble covariance matrix. This points to the need for a restricted version of the notion of an EnKF. We have established a set of necessary and sufficient conditions for the scheme to be unbiased. Whilst these conditions are not a cure-all and cannot deal with independent sources of bias such as modelling errors, they should be useful to designers of EnKFs in the future.

/notices/events/abstracts/stochastic-analysis/mt06/dance.shtml

 

 

http://www.maths.ox.ac.uk/arg

High-performance computing is an important tool for computational science.

Oxford University has recently decided to invest £3M in a new supercomputing

facility which is under development now. In this seminar I will give an overview

of some activities in Oxford and provide a vision for supercomputing here.

I will discuss some of the numerical analysis software and tools,

such as Distributed Matlab and indicate some of the challenges at

the intersection of numerical analysis and high-performance computing.

The talk will describe recent collaborative work between the speaker and Professor Sanjoy Mitter of MIT on connections between continuous-time nonlinear filtering theory, and nonequilibrium statistical mechanics. The study of nonlinear filters from a (Shannon) information- theoretic viewpoint reveals two flows of information, dubbed 'supply' and 'dissipation'. These characterise, in a dynamic way, the dependencies between the past, present and future of the signal and observation processes. In addition, signal and nonlinear filter processes exhibit a number of symmetries, (in particular they are jointly and marginally Markov), and these allow the construction of dual filtering problems by time reversal. The information supply and dissipation processes of a dual problem have rates equal to those of the original, but with supply and dissipation exchanging roles. The joint (signal-filter) process of a nonlinear filtering problem is unusual among Markov processes in that it exhibits one-way flows of information between components. The concept of entropy flow in the stationary distribution of a Markov process is at the heart of a modern theory of nonequilibrium statistical mechanics, based on stochastic dynamics. In this, a rate of entropy flow is defined by means of time averages of stationary ergodic processes. Such a definition is inadequate in the dynamic theory of nonlinear filtering. Instead a rate of entropy production can be defined, which is based on only the (current) local characteristics of the Markov process. This can be thought of as an 'entropic derivative'. The rate of entropy production of the joint process of a nonlinear filtering problem contains an 'interactive' component equal to the sum of the information supply and dissipation rates. These connections between nonlinear filtering and statistical mechanics allow a certain degree of cross- fertilisation between the fields. For example, the nonlinear filter, viewed as a statistical mechanical system, is a type of perpetual motion machine, and provides a precise quantitative example of Landauer's Principle. On the other hand, the theory of dissipative statistical mechanical systems can be brought to bear on the study of sub-optimal filters. On a more philosophical level, we might ask what a nonlinear filter can tell us about the direction of thermodynamic time.