Theatre B, Richard Berry Building
The University of Melbourne
Professor Michael Barnsley
Department of Mathematics and Statistics
The University of Melbourne
It is shown that the invariant measure of a stationary nonatomic stochastic process may be associated both with an Iterated Function System with probabilities and with a certain dynamical system. The theory provides a framework for information theory, and yields a geometrical approach to the discovery of lossless data compression algorithms. As an example, the transform of Burrows and Wheeler is derived. Finally, an open problem is described.