The modeling of various physical questions in plasma kinetics and heat conduction lead to nonlinear boundary value problems involving a nonlocal operator, such as the integral of the unknown solution, which depends on the entire function in the domain rather than at a single point. This talk concerns a particular nonlocal boundary value problem recently studied in [1] by J.R.Cannon and D.J.Galiffa, who proposed a numerical method based on an interval-halving scheme. Starting from their results, we provide a more general convergence theorem and suggest a different iterative procedure to handle the nonlinearity of the discretized problem. References: [1] J.R.Cannon, D.J.Galiffa (2011) On a numerical method for a homogeneous, nonlinear, nonlocal, elliptic boundary problem, Nonlinear Analysis, Vol. 74, pp. 1702-1713.

We consider a weakly nonlinear system of the form (I + phi(x)A)x = p, where phi(x) is a real function of the unknown vector x, and (I + phi(x)A) is an M-matrix. We propose to solve it by means of a sequence of linear systems defined by the iteration procedure (I + phi(x(r))A)x(r + 1) = p, r = 0, 1, ... . The global convergence is proved by considering a related fixed-point problem.

We design and analyse a numerical method for the solution of a particular second order integro-differential boundary value problem on the semiaxis, which arises in the study of the kinetic theory of dusty plasmas. The method we propose represents a first insight into the numerical solution of more complicated problems and consists of a discretization of the differential and integral terms and of an iteration process to solve the resulting non-linear system. Under suitable hypotheses we prove the convergence. We will show the characteristics of the method by means of some numerical simulations.

Starting from the function values on the roots of Jacobi polynomials, we construct a class of discrete de la Vallée Poussin means, by approximating the Fourier coefficients with a Gauss-Jacobi quadrature rule. Unlike the Lagrange interpolation polynomials, the resulting algebraic polynomials are uniformly convergent in suitable spaces of continuous functions, the order of convergence being comparable with the best polynomial approximation. Moreover, in the four Chebyshev cases the discrete de la Vallée Poussin means share the Lagrange interpolation property, which allows us to reduce the computational cost.

We state some pointwise estimates for the rate of weighted approximation of a continuous function on the semiaxis by polynomials. Furthermore we derive matching converse results and estimates involving the derivatives of the approximating polynomials. Using special weighted moduli of continuity, we bridge the gap between an old result by V.M. Fedorov based on the ordinary modulus of smoothness, and the recent norm estimates implicating the Ditzian-Toytik modulus of continuity.

The paper tackles the problem of approximately reconstructing a real function defined on the surface of the unit sphere in the Euclidean q-dimensional space, with q>1, starting from function's samples at scattered sites. Two new operators are introduced for continuous and discrete approximation at scattered sites. Moreover precise error estimates as well as Marcinkiewicz-Zygmund inequalities are derived in every Lp space, giving concrete bounds for all the involved constants.

De la Vallée Poussin means are used to prove Jackson-Favard type estimates for weighted algebraic polynomials with Jacobi and Laguerre-like weights.

In studying some related topics, the authors came back to the paper "On the boundedness of de la Vallée Poussin operators'' [East J. Approx. 7 (2001), no. 4, 417--444] and realized a mistake in the proof of Theorem 2.2. In this note we state the same theorem but slightly modifying the hypothesis on the involved weights.

Starting from a natural generalization of the trigonometric case, we construct a de la Vall\'ee Poussin approximation process in the uniform and L1 norms. With respect to the classical approach we obtain the convergence for a wider class of Jacobi weights. Even if we only consider the Jacobi case, our construction is very general and can be extended to other classes of weights.

The paper gives a contribution of wavelet aspects to classical algebraic polynomial approximation theory. Algebraic polynomial interpolating scaling functions and wavelets are constructed by using the interpolating properties of de la Vallée Poussin kernels w.r.t. the four kinds of Chebyshev weights. For the decomposition and reconstruction of a given function the structure of the involved matrices is studied in order to reduce the computational effort by means of fast cosine and sine transforms.