Time: 1.05 -- 2.00pm;

Room: JCMB 6206.

In this seminar, speakers from Edinburgh ** explain ** their research to a general mathematical audience including final year undergraduate students,
PhD students, postdocs and faculty from the School of Mathematics. We'll try to have a good mix of speakers coming from all the different research areas within the school.
We'll typically have two talks of 25 minutes each.

Please contact the organisers if you are interested in giving a talk in the seminar

(m (dot) kalck (at) ed (dot) ac (dot) uk ; m (dot) lanini (at) ed (dot) ac (dot) uk; k (dot) uda (at) ed (dot) ac (dot) uk.

Mairi Walker (our Mathematics Engagement Officer) kindly offered support in preparing the talks.

Monday October 10, 2016.

13.05 - 13.30: Jacques Vanneste, * Urban pollution and large deviations.*.

13.35 - 14.00: Peter Richtárik, *Stochastic reformulations of linear systems and fast stochastic algorithms.*

Monday October 24, 2016, (13.10 -- 13.45).

13.10 - 13.40: Chris Sangwin, *The pendulum plane and puzzling?*

Monday November 7, 2016.

13.05 - 13.30: Jacques Fleuriot (School of Informatics), *Interactive Theorem Proving = Mathematics ∩ Programming *.

Abstract: * In this talk, I will give a gentle introduction to computer-based
interactive theorem proving and present some arguments as to why it is
a rewarding activity that can blend the joys of mathematics and programming. *

13.35 - 14.00: Ruth King, *Is your model a hidden Markov model? (and why does it matter?)*.

Abstract: * Data are becoming increasingly complex as the quantity and quality
of data is continually increasing. In order to make sense of the data, and increase
our understanding of the underlying system, statistical models are often constructed
to represent the main underlying processes of the system under study. The models are
then fitted to the observed data which in turn provides insight into the system. I will
consider data that are recorded over a period of time, for example, stock market data,
health data (EEG monitoring) or GPS tracking data. For such data the observations are
not independent of each other but are serially correlated (i.e. the observation at time t
is related to the observation at time t-1). In this talk I will present the idea of hidden
Markov models (HMM), demonstrating their versatility to different systems and the
simplifications that can be obtained if we are able to express a model in HMM form.
*

* Monday November 21, 2016.
*

*
*

* 13.10- 13.40: Ben Leimkuhler, Symplectic integrators for fun and profit..
*

*
Abstract: Symplectic integrators are numerical methods which `exactly’ preserve *

an underpinning geometric structure associated to Hamiltonian systems. It has been

found that these elegant schemes allow more faithful reproduction of qualitative

properties and, for example, they are the go-to tool for simulating gravitational systems

(e.g. the future of the solar system) and to simulate trajectories in particle accelerators.

What is perhaps surprising is their profound influence on modern statistics and data science.

I will discuss some basic features of symplectic integrators and attempt to motivate

their widespread popularity.

* See GAMES in autumn 2015 and GAMES in spring 2016 for information about the seminar in previous terms including slides, notes and a video!
*

* Organisers:
Martin Kalck,
Martina Lanini.
Kenneth Uda.
*