Information geometry: optimal experiment design and other applications


Information geometry: optimal experiment design and other applications

Room 215
Peter Hall Building


More information

T: 83446796

The concept of information is useful in physical problems which involve measurement and uncertainty or otherwise naturally contain stochasticity. For example, information quantifies how much one can learn' from a set of noisy measurements drawn from a network of sensors. The mathematical theory of information geometry allows one to use geometric tools to study information in a general setting. The various theorems and constructions of Riemannian geometry can then be used to, for instance, theoretically extremise the information and determine an optimal experimental tuning for a sensor model. In this talk, we introduce the ideas of information geometry with particular emphasis on thetheory of experiments', and discuss some recent mathematical results in the field. We consider a concrete example problem of estimating the parameters of a `target' from noisy bearings measurements, i.e. placing sensors to best estimate the position of some object. Some applications of information geometry to general relativity, thermodynamics, and game theory are also discussed.


  • Mr Arthur Suvorov
    Mr Arthur Suvorov, The University of Melbourne