We will review a Breiman type theorem for Gibbs measures due to Gurevich and Tempelman. For a translation invariant Gibbs measure on a suitable translation invariant configuration set X \subset S^G, where G is an amenable group and S is a finite set, we will prove the convergence of the Shannon-McMillan-Breiman ratio on a specific subset of "generic" configurations. Provided that the above Gibbs measure exists, we also prove the convergence in the definition of pressure and the fact that this Gibbs measure is an equilibrium state.
An M-dependent process X(n) on the integers, is a process for which every event concerning with X(-1),X(-2),... is independent from every event concerning with X(M),X(M+1),...
Such processes play an important role both as scaling limits of physical systems and as a tool in approximating other processes.
A question that has risen independently in several contexts is:
"is there an M dependent proper colouring of the integer lattice for some finite M?"
Abstract: It was noticed in the 30's by Doeblin & Forte that Markov
operators with "chains with complete connections"
act quasi-compactly on the Lipschitz functions. These are operators
like the transfer operators of certain expanding
C^2 interval maps (e.g. the square of Gauss map).
It is folklore that stochastic processes generated by smooth
observables under these maps satisfy many of the results
of "classical probability theory" (e.g. CLT, Chernoff inequality).
I'll try to explain some of this in a "lunchtime" mode.
This talk is devoted to the "Kac-Rice formula", which is an explicit way to compute
the expected number of zeroes of a random series with independent Gaussian coefficients.
We will discuss the original proofs of Kac and Rice (1940's),
an elegant geometrical proof due to Edelman and Kostlan (1995), some interesting examples,
and extensions to complex zeroes and eigenvalues of random matrices.
