Analysis

Block conjugate gradient algorithms for linear systems with multiple right-hand sides

The block conjugate gradient algorithm was proposed in 1980 to solve linear systems with multiple right-hand sides. However, some of the algorithm block vectors may become rank-deficient, leading to numerical difficulties and breakdowns. In this talk, we will study a variant introduced by Dubrulle in 2001 to avoid these problems. We recently proved that this algorithm cannot break down and always converges to the solutions.

On special positive solutions of the first discrete Painlevé hierarchy

The recurrence coefficients of a generalised symmetric Freud weight are positive solutions of a discrete equation in the discrete Painlevé I hierarchy. They also satisfy a coupled system of nonlinear differential equations. Such orthogonality weights also arise in the context of Hermitian matrix models and random symmetric matrix ensembles. In this talk I will report on properties of such special solutions of this integrable system of equations in the dP-I hierarchy, explaining the connections to other areas of mathematics.

A projection based regularized approximation method for ill-posed operator equations

The problem of solving Fredholm integral equations of the first kind is a prototype of an ill-posed problem of the form $T(x) =y$, where $T$ is a compact operator between Hilbert spaces. Regularization and discretization of such equations is necessary for obtaining stable approximate solutions for such problems.  For ill-posed integral equations, a quadrature based collocation method has been considered by Nair (2012) for obtaining discrete regularized approximations.  As a generalization, a projection collocation method has been proposed by the author in 2016.

Resultados de existência global para equações parabólicas com o operador p-Laplaciano.

Resumo: Nesta palestra apresentaremos condições para assegurar a existência global das soluções fraca limitadas do problema de Cauchy da equação parabólica com o termo difusivo dado pelo operador p-Laplaciano (p>2) e condição inicial u_0\in L^\infty (\R^m) \cap L^1(\R^n). Considera-se o caso em que o termo advectivo da equação estimula o crescimento da solução em certas regiões (ou mesmo no espaço todo), de modo a competir com a tendência de decaimento devido ao termo difusivo.