Stochastic gradient descent remains popular in large-scale machine learning, on account of its very low computational cost and robustness to noise. However, gradient descent is only linearly efficient and not transformation invariant. Scaling by a local measure can substantially improve its performance. One natural choice of such a scale is the Hessian of the objective function: Were it availab...

The Jacobi matrix of the (2n+1)-point Gauss-Kronrod quadrature rule for a given measure is calculated efficiently by a five-term recurrence relation. The algorithm uses only rational operations and is therefore also useful for obtaining the Jacobi-Kronrod matrix analytically. The nodes and weights can then be computed directly by standard software for Gaussian quadrature formulas.

In this paper, we propose and analyze a new predictor-corrector method for solving a nonlinear system of equations using the weight combination of mid-point, Trapezoidal and quadrature formulas. A new algorithm based on this method has been presented. The proposed method has been tested on a series of examples published in the literature and shows good results. Comparison with other similar met...

Suppose that the inverse scattering problem is understood as follows: given fixed-energy phase shifts, corresponding to an unknown potential q = q(r) from a certain class, for example, q ∈ L 1,1 , recover this potential. Then it is proved that the Newton-Sabatier (NS) procedure does not solve the above problem. It is not a valid inversion method, in the following sense: 1) it is not possible to...

We present a new numerical method, based on a coupling of finite elements and boundary elements, to solve a fluid–solid interaction problem posed in the plane. The boundary unknowns involved in our formulation are approximated by a spectral method. We provide error estimates for the Galerkin method, propose fully discrete schemes based on elementary quadrature formulas and show that the perturb...

Several deenitions of universality of an n-point quadrature formula Q n are discussed. Universality means that Q n is able to compete with the respective optimal formula in many classes of functions. It is proved in a certain sense that the Gaus-sian quadrature formula satisses such a universality criterion. The underlying classes of functions are In each of these classes, we loose at most the ...

In a previous paper it was shown how a probability measure (Eberlein measure) on the closed unit ball of the sequence space, h, can be used to find the variance a2 of the error functional for a quadrature formula for the fc-dimensional cube, regarded as a random variable. Here we give values of a for some specific formulae.