to be done in this area. Face recognition is a problem that scarcely clamoured for attention before the computer age but, having surfaced, has involved a wide range of techniques and has attracted the attention of some fine minds (David Mumford was a Fields Medallist in 1974). This singular application of mathematical modelling to a messy applied problem of obvious utility and importance but with no unique solution is a pretty one to share with students: perhaps, returning to the source of our opening quotation, we may invert Duncan's earlier observation, 'There is an art to find the mind's construction in the face!'.

Linear and nonlinear waves, by G. B. Whitham. Pp.636. £50. 1999. ISBN 0 471 35942 4 (Wiley).

Large linear systems of saddle point type arise in a wide variety of applications throughout computational science and engineering. Due to their indefiniteness and often poor spectral properties, such linear systems represent a significant challenge for solver developers. In recent years there has been a surge of interest in saddle point problems, and numerous solution techniques have been proposed for this type of system. The aim of this paper is to present and discuss a large selection of solution methods for linear systems in saddle point form, with an emphasis on iterative methods for large and sparse problems.

/pdf/numerical-solution-of-saddle-point-problems-l583iob5c2.pdf

Numerical solution of saddle point problems

Finite element methods for Maxwell's equations.

Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD), also called algorithmic differentiation or simply "auto-diff", is a family of techniques similar to but more general than backpropagation for efficiently and accurately evaluating derivatives of numeric functions expressed as computer programs. AD is a small but established field with applications in areas including computational uid dynamics, atmospheric sciences, and engineering design optimization. Until very recently, the fields of machine learning and AD have largely been unaware of each other and, in some cases, have independently discovered each other's results. Despite its relevance, general-purpose AD has been missing from the machine learning toolbox, a situation slowly changing with its ongoing adoption under the names "dynamic computational graphs" and "differentiable programming". We survey the intersection of AD and machine learning, cover applications where AD has direct relevance, and address the main implementation techniques. By precisely defining the main differentiation techniques and their interrelationships, we aim to bring clarity to the usage of the terms "autodiff", "automatic differentiation", and "symbolic differentiation" as these are encountered more and more in machine learning settings.

/pdf/automatic-differentiation-in-machine-learning-a-survey-4s77ht2sqh.pdf

Automatic differentiation in machine learning: a survey

/pdf/numerical-approximation-methods-for-elliptic-boundary-value-4emguz66zy.pdf

Numerical Approximation Methods For Elliptic Boundary Value Problems

In this paper we introduce the Boundary Element Tearing and Interconnecting (BETI) methods as boundary element counterparts of the well-established Finite Element Tearing and Interconnecting (FETI) methods. In some practical important applications such as far field computations, handling of singularities and moving parts etc., BETI methods have certainly some advantages over their finite element counterparts. This claim is especially true for the sparse versions of the BETI preconditioners resp. methods. Moreover, there is an unified framework for coupling, handling, and analyzing both methods. In particular, the FETI methods can benefit from preconditioning components constructed by boundary element techniques. The first numerical results confirm the efficiency and the robustness predicted by our analysis.

Boundary Element Tearing and Interconnecting Methods

This work is devoted to non-linear eddy current problems and their numerical treatment by the so-called multiharmonic approach. Since the sources are usually alternating currents, we propose a truncated Fourier series expansion instead of a costly time-stepping scheme. Moreover, we suggest to introduce some regularization parameter that ensures unique solvability not only in the factor space of divergence-free functions, but also in the whole space H(curl). Finally, we provide a rigorous estimate for the total error that is due to the use of truncated Fourier series, the regularization technique and the spatial finite element discretization.

/pdf/numerical-analysis-of-nonlinear-multiharmonic-eddy-current-5gc94qazrs.pdf

Numerical analysis of nonlinear multiharmonic eddy current problems

https://www.ricam.oeaw.ac.at/files/reports/15/rep15-19.pdf

Space–time isogeometric analysis of parabolic evolution problems

Many surface acoustic wave (SAW) devices consist of quasiperiodic structures that are designed by successive repetition of a base cell. The precise numerical simulation of such devices, including all physical effects, is currently beyond the capacity of high-end computation. Therefore, we have to restrict the numerical analysis to the periodic substructure. By using the finite-element method (FEM), this can be done by introducing periodic boundary conditions (PBCs) at special artificial boundaries. To be able to describe the complete dispersion behavior of waves, including damping effects, the PBC has to be able to model each mode that can be excited within the periodic structure. Therefore, the condition used for the PBCs must hold for each phase and amplitude difference existing at periodic boundaries. Based on the Floquet theorem, our two newly developed PBC algorithms allow the calculation of both, the phase and the amplitude coefficients of the wave. In the first part of this paper we describe the basic theory of the PBCs. Based on the FEM, we develop two different methods that deliver the same results but have totally different numerical properties and, therefore, allow the use of problem-adapted solvers. Further on, we show how to compute the charge distribution of periodic SAW structures with the aid of the new PBCs. In the second part, we compare the measured and simulated dispersion behavior of waves propagating on periodic SAW structures for two different piezoelectric substrates. Then we compare measured and simulated input admittances of structures similar to SAW resonators.

/pdf/finite-element-simulation-of-wave-propagation-in-periodic-1jfxs76kv9.pdf

Finite-element simulation of wave propagation in periodic piezoelectric SAW structures

Isogeometric analysis (IGA) is a recently developed simulation method that allows integration of finite element analysis (FEA) with conventional computer-aided design (CAD) software [1,3]. This goal requires new software design strategies, in order to enable the use of CAD data in the analysis pipeline. To this end, we have initiated G+SMO (Geometry+Simulation Modules), an open-source, C++ library for IGA. G+SMO is an object-oriented, template library, that implements a generic concept for IGA, based on abstract classes for discretization basis, geometry map, assembler, solver and so on. It makes use of object polymorphism and inheritance techniques to provide a common framework for IGA, for a variety of different basis-types which are available. A highlight of our design is the dimension independent code, realized by means of template meta-programming. Some of the features already available include computing with B-spline, Bernstein, NURBS bases, as well as hierarchical and truncated hierarchical bases of arbitrary polynomial order. These basis functions are used in continuous and discontinuous Galerkin approximation of PDEs over (non-)conforming multi-patch computational (physical) domains. (© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)

/pdf/geometry-simulation-modules-implementing-isogeometric-2ivu27rmou.pdf

Ulrich Langer

Papers

Boundary Element Tearing and Interconnecting Methods

Numerical analysis of nonlinear multiharmonic eddy current problems

Space–time isogeometric analysis of parabolic evolution problems

Finite-element simulation of wave propagation in periodic piezoelectric SAW structures

Geometry + Simulation Modules: Implementing Isogeometric Analysis