November 10, 1999, 4.00 pm
Lecture Theatre OF/B9 Oddfellows Hall (Material Science)
Forget about predicting the future, try knowing the present
Dr. Kevin Judd, Dept. of Mathematics &
Statistics , University of
Western Australia
If one has a perfect model of a dynamical system, which is deterministic
and finite dimensional, then the best forcast or prediction of the
future at any moment is attained by determining the state of the system
at that moment. However, if the observation of the state is clouded by
observation error, then one must estimate the state given what has been
observed up to that moment. The question is, can an exact estimation of
the state be got from infinite past observation? The answer will be
shown to be no. This implies that forcasts must take into account the
distribution of the indistiguishable states, for example, by using an
ensemble of states. A new method of estimating the state and ensembles
will be presented that can be used with observational error, with
dynamic noise, with imperfect models, and, when the recurrence time of
the attractor is long.
Back to the Seminar Homepage
For further info contact either Matthias
Heil (mheil@ma.man.ac.uk),
Mark Muldoon
(M.Muldoon@umist.ac.uk)or the seminar secretary (Tel. 0161 275 5800).
Page last modified: October 27, 1999
Back to the
department's home page.