A Stochastic Subgradient Method for Distributionally Robust Non-convex and Non-smooth Learning

Journal of Optimization Theory and Applications - We consider a distributionally robust formulation of stochastic optimization problems arising in statistical learning, where robustness is with...     

Galerkin method for optimal control of second‐order evolution equations

This paper presents the Galerkin approximation of the optimization problem of a system governed by non‐linear second‐order evolution equation where a non‐linear operator depends on derivative of the state of the system. The control is acting on a non‐linear equation. After giving some results on the existence of optimal control we shall prove the existence of the weak and strong condensation points of a set of solutions of the approximate optimization problems. Each of these points is a solution of the initial optimization problem. Finally we shall give a simple example using the obtained results. Copyright © 2004 John Wiley & Sons, Ltd.     

Variety of LaSrMnO structures induced by growth conditions and laser irradiation

Crystallographic phase transitions in perovskite-like LaSrMnO metallic oxides are studied. The transitions are induced when internal stresses generated during film synthesis (at temperatures between 450 and 730°C) vary (decrease or increase) upon subsequent irradiation by a KrF laser emitting in the UV range. As the synthesis temperature T _s grows, the rhombohedral-to-orthorhombic phase transition occurs at 650–670°C. The resistivity is shown to be either temperature-independent, ρ(T)=const, at T<T _crit, or varies and reaches a maximum, ρ(T)=ρ_max, at the Curie temperature T _c. Optical transmission spectra taken at photon energies ℏω=0.5–2.5 eV exhibit both a high (0.8–0.9) and low (0.1–0.3) transmission coefficient t, depending on the synthesis temperature. As follows from X-ray diffraction data, the laser irradiation causes a phase transition only in LaSrMnO films grown at T _s<650°C. Phases of different size scales appear: the long-range-order orthorhombic matrix and mesoscopic-range-order rhombohedral clusters are observed in the films grown at T _s=450–550°C and the rhombohedral matrix with orthorhombic clusters, in the films grown at T _s=550–650°C.     

On testing the divisibility of lacunary polynomials by cyclotomic polynomials

An algorithm is described that determines whether a given polynomial with integer coefficients has a cyclotomic factor. The algorithm is intended to be used for sparse polynomials given as a sequence of coefficient-exponent pairs. A running analysis shows that, for a fixed number of nonzero terms, the algorithm runs in polynomial time.

     

Standardization of Methods for Characterizing the Surface Geometry of Solids

Since a comprehensive survey published in 1999 [1] much work was done in standardizing measuring methods to characterize the surface geometry of dispersed and/or porous solids and to certify reference materials. The present paper is an extension of a short communication [2]. It gives a survey on existing standards and reports on new drafts and proposals.     

The representation dimension of domestic weakly symmetric algebras

Auslander’s representation dimension measures how far a finite dimensional algebra is away from being of finite representation type. In [1], M. Auslander proved that a finite dimensional algebra A is of finite representation type if and only if the representation dimension of A is at most 2. Recently, R. Rouquier proved that there are finite dimensional algebras of an arbitrarily large finite representation dimension. One of the exciting open problems is to show that all finite dimensional algebras of tame representation type have representation dimension at most 3. We prove that this is true for all domestic weakly symmetric algebras over algebraically closed fields having simply connected Galois coverings.     

A marker method for the solution of the damped Burgers' equation

A new method for the solution of the damped Burgers' equation is described. The marker method relies on the definition of a convective field associated with the underlying partial differential equation; the information about the approximate solution is associated with the response of an ensemble of markers to this convective field. Some key aspects of the method, such as the selection of the shape function and the initial loading, are discussed in some details. The marker method is applicable to a general class of nonlinear dispersive partial differential equations. © 2005 Wiley Periodicals, Inc. Numer Methods Partial Differential Eq, 2006