# Past Seminars

Page 1 2 3 4 5 6 7 8 9

Mathematical modelling of macrosystems: development and application of the maximum entropy method and local equilibria

*Speaker:* Yu.S Popkov

*Institution:* Institute for System Analysis of the Russia Academy of Sciences, Moscow

*Date:* Wed 18 Dec 2013

*Time:* 3:15 pm

*Location:* Room 213, Richard Berry Building, University of Melbourne*Abstract:* The talk is a survey of my work on the general entropy approach to system analysis that is based on the theory of macrosystems, and its applications in different fields. The first part of the talk is devoted to the principles of and mathematical tools for building models of stationary and non-stationary states, to studying the parametric properties of the models, and to multiplicative computational methods. Modelling of stationary states is described as the problem of entropy maximization over a feasible set. Models for non-stationary states are described as dynamic systems with entropy operator. The second part of the talk is devoted to applications in socio-economic modeling (integrated demographic and economic models), to technical problems (entropy for image reconstruction from
projections) and data analysis (entropy-robust estimation under limited data).

The Kelly Criterion: the application of information theory to horse racing

*Speaker:* Caley Finn

*Institution:* University of Melbourne

*Date:* Tue 1 Nov 2011

*Time:* 1:00 pm

*Location:* Room 213, Richard Berry Building, University of Melbourne*Abstract:* In his 1956 paper, Kelly imagined a gambler with a secret channel
telling him the outcome of races - with perfect communication, the
gambler could wager his entire bankroll and win every time. However,
if instead this insider information is received through a noisy
channel, with some probability of error, the gambler should bet only a
fraction of his bankroll or face certain ruin. The Kelly criterion
gives the fraction to bet in order to maximise growth of the bankroll.
In this talk, Caley Finn will sketch the derivation of the Kelly criteria, tell
you how it will make you rich (Thorpe, 1966), why it will plunge you
into depression (Leib, 2000), and why it is of no use for an
occasional bet on the horses. Caley will then apply it (inappropriately)
to tell you how to spend your money on the Melbourne Cup.

Adventures in Bayesian Statistics

*Speaker:* Professor Kerrie Mengersen

*Institution:* Queensland University of Technology

*Date:* Tue 25 Oct 2011

*Time:* 5:45 pm

*Location:* J H Michell Theatre, Richard Berry Building, Department of Mathematics & Statistics*Abstract:* The Maurice H. Belz Lecture.

Thinking about sampling variability and replication: Confidence intervals beat p values

*Speaker:* Emeritus Professor Geoff Cumming

*Institution:* Statistical Cognition Laboratory, School of Psychological Science, La Trobe University

*Date:* Fri 21 Oct 2011

*Time:* 12:00 pm

*Location:* Room 310 (Access Grid Room), Physical Sciences 2, La Trobe University, Bundoora Campus*Abstract:* Understanding sampling variability is an important goal. A useful approach is to consider replication. What information do inferential techniques give about replication? Confidence intervals (CIs) do well: A 95% CI is, approximately, an 83% prediction interval for the mean of a replication experiment. A p value, however, gives almost no information about a replication result. CIs beat p values, although for very small N a CI may be misleading. Professor Cumming will present novel graphics and simulations, and emphasise cognition - how people read graphics and draw conclusions. There is more in his book: 'Understanding The New Statistics: Effect Sizes, Confidence Intervals, and Meta-Analysis', Routledge, 2012. www.thenewstatistics.com.
Contact: Dr Andriy Olenko;
Statistics Seminar Convenor, Department of Mathematics and Statistics,
a.olenko@latrobe.edu.au

Forecasting Model Validation

*Speaker:* Julia Polak

*Institution:* Monash University

*Date:* Fri 7 Oct 2011

*Time:* 11:00 am

*Location:* Room 310 (Access Grid Room), Physical Sciences 2, La Trobe University, Bundoora*Abstract:* Forecasting models play a crucial role in many decision-making areas. Many tools have been developed for model selection and validation (on available data) but only a few exist for answering the question whether the model under the test is still valid for the new observations. This is especially true when one is looking for a quick answer after a small number of extra observations have become available or examining a nonparametric model. We present a method for analyzing the model, which has already been selected, and examine whether its predictive ability is still good enough or the model needs to be reworked. The proposed prediction capability procedure combines the ideas of nonparametric density estimation and principal function data analysis in order to clarify the question whether the new observed data comes from the same expected data generation process or not. If there is not enough evidence that the data generation process has been changed after the model has been selected, there is no reason to believe that the model lost its predictive abilities in the new reality.
Contact: Dr Andriy Olenko:
a.olenko@latrobe.edu.au

HPC Forum at University of Melbourne

*Speaker:* Kimberly Powell

*Institution:*

*Date:* Thu 29 Sep 2011

*Time:* 3:00 pm

*Location:* The University of Melbourne, Parkville Campus 207 Bouverie Street, Theatre 1 (entrance at 221 Bouverie Street)*Abstract:* 3.00-4.00pm: Kimberly Powell, Life Sciences specialist from NVIDIA, will be sharing her expertise in developing high performance computing approaches to addressing the world's most challenging Life Sciences and Medical Imaging research problems.
4.00 – 5.00pm: tea/coffee/discussion.
No RSVP. If you require further information, please send an email to Forum Coordinator Sara Ogston, sogston@unimelb.edu.au.
The University of Melbourne ITS Research Services, IBM Research Collaboratory for Life Sciences and the Victorian Life Sciences Computation Initiative (VLSCI) formed the HPC Forum to connect Victorian researchers and students with an interest in high performance computation. Regular meetings are usually held monthly, with additional special events.

How to reconstruct a metric by its unparameterised geodesics

*Speaker:* Prof Vladimir Matveev

*Institution:* University of Jena, Germany

*Date:* Mon 26 Sep 2011

*Time:* 2:00 pm

*Location:* Room 310 (Access Grid Room), Physical Sciences 2, La Trobe University, Bundoora*Abstract:* Prof Vladimir Matveev will discuss whether it is possible to reconstruct a metric by its unparameterized geodesics, and how to do it effectively. Prof Matveev explains why this problem is interesting for general relativity, shows how to understand whether all curves from a sufficiently big family are unparameterized geodesics of a certain affine connection, and how to reconstruct algorithmically a generic 4-dimensional metric by its unparameterised geodesics. The algorithm works most effectively if the metric is Ricci-flat. Prof Matveev will also prove that almost every metric does not allow nontrivial geodesic equivalence, and construct all pairs of 4-dimensional geodesically equivalent metrics of Lorenz signature. If the time allows, Prof Matveev will also explain how this theory helped to solve two problems explicitly formulated by Sophus Lie in 1882, and the semi-Riemannian two-dimensional version of the projective Lichnerowicz-Obata conjecture. The new results of the talk are based on the papers arXiv:1010.4699, arXiv:1002.3934, arXiv:0806.3169, arXiv:0802.2344 and arXiv:0705.3592; joint with Bryant, Bolsinov, Kiosak, Manno and Pucacco.
Contact:
Dr Yury Nikolayevsky;
Seminar Convenor, Department of Mathematics and Statistics,
La Trobe University:
Y.Nikolayevsky@latrobe.edu.au

Geodesically equivalent metrics: on the crossroad of differential geometry, integrable systems and mathematical physics.

*Speaker:* Prof Vladimir Matveev

*Institution:* University of Jena, Germany

*Date:* Mon 26 Sep 2011

*Time:* 1:00 pm

*Location:* Room 310 (Access Grid Room), Physical Sciences 2, La Trobe University, Bundoora*Abstract:* Can two different metrics have the same geodesics? Yes! The first examples were constructed already by Lagrange, and different versions of the question were actively studied by virtually all differential geometers 100 years ago. During his talk Prof Matveev will explain the solution of the Lie Problem which is the infinitesimal version of the question above (this is a joint result with R. Bryant and G. Manno), of the Beltrami Problem (which is precisely the question above, Prof Matveev's contribution is to solve it on closed manifolds), and of the Lichnerowicz-Obata conjecture (which suggests an answer to the Schouten problem). There are three main tools of the proof: integrable systems, geometric theory of partial differential equations and singularity theory.
Contact:
Dr Yury Nikolayevsky;
Seminar Convenor, Department of Mathematics and Statistics,
La Trobe University:
Y.Nikolayevsky@latrobe.edu.au

La Trobe Mathematics Colloquium: Dirichlet analogues of q-series and some tiling patterns from quasicrystals

*Speaker:* Dr Geoffrey Campbell

*Institution:* La Trobe University

*Date:* Fri 2 Sep 2011

*Time:* 2:00 pm

*Location:* Access Grid Room (Room 310, PS2)*Abstract:* In writing new analogues for classical q series connected with Dirichlet series generating functions, the issue of natural interpretation of coefficients arose. Happily, many such interpretations have recently arisen in enumerative tilings that occur in the theory of quasicrystals. To be able to explain the concepts in a simple fashion, the speaker will mainly use two examples, discussed previously by Baake and Grimm, namely the triangular lattice, the set (ring) of Eisenstein integers, and the vertex set of the twelvefold symmetric shield tiling. The speaker will concentrate on explicit results for these examples, and refer to original sources for a more general exposition and for details.

What is constraint integer programming?

*Speaker:* Timo Berthold

*Institution:* Zuse Institute Berlin and Matheon

*Date:* Fri 29 Oct 2010

*Time:* 4:15 pm

*Location:* Alice Hoy, Room 307, Floor 3*Abstract:* Mixed-integer programming (MIP) and constraint programming (CP) proved to
be a powerful tools to model and solve
large-scale optimization problems. Constraint integer programming (CIP) is
a novel generalization of MIP that supports
the notion of arbitrary constraints as in CP. We introduce the basic
notion and algorithmic ideas of CIP. Further, we
present the software SCIP which is a solver and framework for constraint
integer programming that also features SAT
solving techniques. SCIP is available in source code and free for
non-commercial use.

We illustrate the algorithmic design and the main sequence of the solving
steps. Furthermore, we describe the various
algorithmic components that enrich the basic CIP framework and discuss
their role in the solving process. In this talk,
we will mainly focus on techniques for solving mixed-integer programs.
Computational experiments indicating the
potential of the approach are provided.

Bio: Timo is one of the chief developers of SCIP (an integer programming
solving framework developed at Zuse Institute Berlin). He is an expert on
primal heuristics and branching rules for mixed-integer linear and
nonlinear programs.

Page 1 2 3 4 5 6 7 8 9