Georgia Institute of Technology College of Sciences

A fundamental result in the geometry of numbers states that any lattice free convex set in R^n has integer width bounded by a function of dimension, i.e. the so called Flatness Theorem for Convex Bodies. This result provides the theoretical basis for the polynomial solvability of Integer Programs with a fixed number of (general) integer variables. In this work, we provide a simplified proof of the Flatness Theorem with tighter constants. Our main technical contribution is a new tight bound on the smoothing parameter of a lattice, a concept developed within lattice based cryptography which enables comparisons between certain discrete distributions over integer points with associated continuous Gaussian distributions. Based on joint work with Kai-Min Chung, Feng Hao Liu, and Christopher Peikert.

We study the nonparametric regression model (X1 , Y1 ), ...(Xn , Yn ) , where (Xi )i≥1 is the deterministic design and (Yi )i≥1 is a sequence of real random variables. Assume that the density of Yi is known and can be written as g (., f (Xi )) , which depends on a regression function f at the point Xi . The function f is assumed smooth, i.e. belonging to a Hoelder ball or a Nikol’ski ball. The aim is to estimate the regression function from the observations for two error risks (pointwise and global estimations) and to find the optimal estimator (in the sense of rates of convergence) for each density g . We are particularly interested in the study of irregular models, i.e. when the Fisher information does not exist (for example, when the density g is discontinuous like the uniform density). In this case, the rate of convergence can be improved with the use of nolinear estimators like Maximum likelihood or bayesian estimators. We use the locally parametric approach to construct a new local version of bayesian estimators. Under some conditions on the likelihood of the model, we propose an adaptive procedure based on the so-called Lepski’s method (adaptive selection of the bandwidth) which allows us to construct an optimal adaptive bayesian estimator. We apply this theory to several models like multiplicative uniform model, shifted exponential model, alpha model, inhomogeous Poisson model and Gaussian model

Binary function is a class of important function that appears in many applications e.g. image segmentation, bar code recognition, shape detection and so on. Most studies on reconstruction of binary function are based on the nonconvex double-well potential or total variation. In this research we proved that under certain conditions the binary function can be reconstructed from incomplete frequency information by using only simple linear programming, which is far more efficient.

We present a novel method to find KAM tori in degenerate (nontwist) cases. We also require that the tori thus constructed have a singular Birkhoff normal form. The method provides a natural classification of KAM tori which is based on Singularity Theory.The method also leads to effective algorithms of computation, and we present some preliminary numerical results. This work is in collaboration with R. de la Llave and A. Gonzalez.