Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Convergence of Nearest Neighbor Markov chains on discrete trees towards Brownian Motion on real-trees Siva Athreya Indian Statistical Institute, Bangalore We will begin with a review of the ν-Brownian motion on any locally compact R-tree (T, r) equipped with a Radon measure. Then we will show that this ν-Brownian motion on (T, r) is the scaling limit of nearest neighbor Markov chains on discrete metric graphs (Tn , rn ) which jump from a vertex v ∈ Tn to a neighboring vertex v 0 ∼ v with rate (νn ({v}) · rn (v, v 0 ))−1 provided that (Tn , rn , νn ) converges to (T, r, ν) in a Gromov-weak type of topology. We shall explain the corresponding topology as well. This is joint work with Wolfgang Lohr and Anita Winter A stability theorem for the elliptic Harnack inequality Richard Bass University of Connecticut Harnack inequalities are an important tool in probability theory, as well as for analysis and partial differential equations. I will talk about a stability theorem for the elliptic Harnack inequality: if two weighted graphs are equivalent, then the elliptic Harnack inequality holds for harmonic functions with respect to one of the graphs if and only if it holds for harmonic functions with respect to the other graph. I’ll also give a characterization of the elliptic Harnack inequality. Limiting spectral distribution of patterned random matrices Arup Bose Indian Statistical Institute, Kolkata We present a unified approach to establishing limiting spectral distribution (LSD) of patterned matrices via the moment method. This allows us to demonstrate relatively short proofs for the LSDs of common matrices (Wigner, Toeplitz, Hankel, Reverse Circulant, Symmetric Circulant) and provide insight into the nature of different LSDs and their interrelations. The method is applicable to matrices with appropriate dependent entries, banded matrices (including triangular matrices) and matrices of the form Ap = n1 XX 0 where X is a p × n matrix with real entries and p → ∞ with n = n(p) → ∞ and p/n → y with 0 ≤ y < ∞. The sample 1 variance covariance matrix being a particular example of the latter. We can also establish the existence of the LSD of the sample autocovariance matrix by this method. Joint convergence of several copies of different patterned matrices may also be established by this approach. This approach raises interesting questions about the class of patterns for which LSD exists and the nature of the possible limits. In many cases the LSDs are not known in any explicit forms and so deriving probabilistic properties of the limit are also interesting issues. Certain small noise limits for diffusions Vivek Borkar Indian Institute of Technology, Mumbai This talk will describe some recent work (joint with K. Sureshkumar) on the small noise large time asymptotics of the normalized Feynman-Kac semigroup using control theoretic techniques. Pathwise uniqueness of stochastic partial differential equations in Banach spaces. Sandra Cerrai University of Maryland We prove pathwise uniqueness for an abstract stochastic reaction-diffusion equation in Banach spaces. The drift contains a bounded Holder term; in spite of this, due to the space-time white noise it is possible to prove pathwise uniqueness. The proof is based on a detailed analysis of the associated Kolmogorov equation. The model includes examples not covered by the previous works based on Hilbert spaces or concrete SPDEs. This is a joint work with G. da Prato and F. Flandoli. Conditional ergodicity in infinite dimension Ramon Van Handel Princeton University The classical ergodic theory of Markov chains has achieved a rather definitive form in the theory of Harris chains, which characterizes total variation convergence to equilibrium. While this theory is formulated in principle for any measurable state space, it is mainly applicable in finite dimension: transition probabilities in infinite dimension are typically singular, so that the total 2 variation convergence cannot hold. In models such as stochastic PDEs, ergodicity has therefore largely been investigated by topological (rather than measure-theoretic) methods. While many beautiful results can be obtained in this manner, topological methods are not well suited to certain probabilistic questions. We are particularly motivated by the problem of establishing ergodicity of nonlinear filters when the underlying model is a stochastic PDE: as conditioning is a very singular measure-theoretic operation, topological properties are not expected to be preserved. Nonetheless, I will argue that topological ergodicity can often be encoded in a completely measure-theoretic fashion. This surprisingly simple idea provides a key tool for establishing conditional ergodicity in infinite dimension. (Joint work with Xin Thomson Tong.) Nodal length of random eigenfunctions of the Laplacian on the 2-d torus Manjunath Krishnapur Indian Institute of Science, Bangalore We consider Gaussian linear combinations of eigenfunctions of the Laplacian on the 2-dimensional torus for a given eigenvalue. The mean of the length of the nodal set is easy to compute. As the dimension of the eigenspace goes to infinity, we find asymptotics of the variance of the nodal length. The proof has two main steps. First, the Kac-Rice formulas from probability give a formula for the variance in terms of the covariance kernel of the Gaussian random function. The covariance kernel is of an arithmetic nature and is analyzed using tools from additive combinatorics. This is joint work with Igor Wigman and Pr Kurlberg. Time-change equations for diffusion processes Tom Kurtz University of Wisconsin General notions of weak and strong solutions of stochastic equations will be described and a general version of the Yamada-Watanabe-Engelbert theorem relating existence and uniqueness of weak and strong solutions given. Time-change equations for diffusion processes provide an interesting example. Such equations arise naturally as limits of analogous equations for Markov chains. For one-dimensional diffusions they also are essentially given in the now-famous notebook of Doeblin. Although requiring nothing more than standard Brownian motions and the Riemann integral to define, the question of strong uniqueness remains unresolved. To prove weak uniqueness, the notion of compatible solution is introduced and the martingale properties of compatible solutions used to reduce the uniqueness question to the corresponding question for a martingale problem or an Ito equation. 3 Rates of convergence of color count in balanced urn models Krishanu Maulik Indian Statistical Institute, Kolkata We consider urn models containing balls of finitely many colors and with balanced replacement matrices, that is, replacement matrices with common row sum. We study the rates of convergence of each color count and provide relationship between the limiting random variables corresponding to various colors thus obtained. We show that the convergence occurs both in almost sure and in Lp sense, for any p > 0. Parts of the work are jointly done with Amites Dasgupta and Gourab Ray. Brownian Web in the Scaling Limit of Supercritical Oriented Percolation in Dimension 1 + 1 Anish Sarkar Indian Statistical Institute, Kolkata In this talk, I will discuss the connection of oriented percolation with Brownian Web. In particular, we show that, after centreing and diffusively rescaling space and time, the collection of rightmost infinite open paths in a supercritical oriented percolation configuration on the spacetime lattice Z2even := {(x, i) ∈ Z2 : x + i is even} converges in distribution to the Brownian web. This proves a conjecture of Wu and Zhang (2008). Our key observation is that each rightmost infinite open path can be approximated by a percolation exploration cluster, and different exploration clusters evolve independently before they intersect. A metric between probability distributions of different sizes and applications to order reduction Mathukumalli Vidyasagar University of Texas at Dallas In this talk we define a metric distance between probability distributions when they are defined over finite sets of different cardinalities. The metric is obtained by finding a joint distribution whose marginals are the two given distributions, such that its entropy is minimized. This is equivalent to maximizing the mutual information between two random variables having the two given distributions. It turns out that in general actually computing the distance is NP-hard, so a greedy algorithm is proposed instead, based on viewing this as a bin-packing problem. Next, the problem of optimally approximating a high-order distribution by a lower-order distribution (in this metric) is studied. It is shown that the optimal approximation is always an aggregation of the original distribution that has maximum entropy (amongst all aggregations). Finding the optimal aggregation also turns out to be NP-hard, so this problem is also formulated as a bin-packing problem and a greedy algorithm is proposed for its solution. 4