The 27th Biennial Conference on Numerical Analysis

The 27th Biennial Conference in Numerical Analysis took place at the University of Strathclyde, 27-30 June.

The series began at St. Andrews University in 1965, then spent many happy years at the University of Dundee, and finally moved to Glasgow in 2009. A full history, up to 2005, can be found on Alistair Watson’s website, and comments on the 50th Anniversary Meeting can be found on Nick Higham’s blog.

Numerical Analysis pens
Pens from the 20th, 21st, and 22nd biennial conferences on Numerical Analysis.

The conference has a traditional feel in that, whereas many meetings of its scale are held at hotels and conference centres, this one is entirely organised by academics, and hosted on a university campus. This traditional feel contrasts with a modern focus: through its twelve plenary talks, fifteen mini-symposia, and numerous contributed talks, the Biennial Conference reflects the many recent innovations in numerical analysis.

The plenary talks are at the core of this meeting, which opened with Christian Lubich (Tubingen) presenting recent developments in dynamic low-rank approximation, and Valeria Simoncini’s talk on methods for large-scale Sylvester equations (and related problems). David Keyes kept the focus on (extremely) large-scale problems, exhorting the numerical analysis community to develop algorithms that keep pace with modern hardware. The (now) classical roots of numerical analysis, and their influence on current trends were the focus of, for example, the Fletcher-Powell lecture, given by Philip Gill (San Diego). And that was all on Day 1, which concluded with a civic reception hosted by The City of Glasgow in the spectacular City Chambers.

 

DEstep
Donald Estep (Colorado) making the case for computational measure theory

Wednesday’s plenaries were delivered by Donald Estep (Colorado), Ilaria Perugia (Vienna), and Gerlind Plonka-Hoch on, respectively, “computational measure theory” (with applications to modelling extreme weather events), completely discontinuous finite elements, and sublinear sparse FFT methods.

AnnaKarinTornberg
Anna-Karin Tornberg (KTH) on Quadrature by expansion (source: Heather Yorston)

Anna-Karin Tornberg (KTH; above) opened Thursday’s sessions. She was followed by Endre Suli (Oxford) who married the classical and modern, where he outlined the analysis of finite element approximations for viscous incompressible fluids, accompanied by a live demonstration with tooth-paste.

Through out the meeting, it is clear that the distinctions between numerical analysis and computer science are becoming increasingly blurred, particularly in the field of networks, where algorithms for graphs and matrices are of core importance. This featured in the A.R. Mitchell lecture, delivered by Andrew Stuart (Caltech). He discussed machine learning and for classification algorithms applied on large graphs.

That evening the conference dinner was hosted in the remarkable Trades Hall, which was designed and built between 1791 and 1794. The after dinner speech, which was both humorous and thoughtful,  was given by Ivan Graham (Bath).

TopTable
View from the top table (source: David Gleich)

Arguably, the best was saved until last. In Friday morning’s first plenary lecture, Francoise Tisseur (Manchester) made a compelling case for the use of tropical algebra in
numerical analysis, with the applications to incomplete factorisations being particularly compelling. She was followed by David Gleich (Purdue), who explained the motivation and methods for locating motifs (such as triangles) in networks.

Of course, there was much more to the conference than the plenary talks and receptions. There were 15 minisymposia on topics ranging from “M1” on non-local problems (such as the en vogue fractional differential equations) to “M15” on Chebfun, as well as contributed talks. In particular, there were numerous presentations by student participants. Three of these were awarded prizes by the UK and Ireland section of SIAM:

  • Denis Devaud, ETH Zurich, Switzerland, “Exponential convergence in
    H^1/2 of hp-approximation for parabolic equations”
  • James Rynn, University of Manchester, UK, “Using Surrogate Models to
    Accelerate Bayesian Inverse Uncertainty Quantification”
  • Florian Wechsung, University of Oxford, UK, “Shape Optimization with
    Geometric Constraints Using Moreau-Yosida Regularization”

The judging panel was Ivan Graham (Bath), Natalia Kopteva  (Limerick) and John Mackenzie (Strathclyde).

To conclude, the 27th conference was at least as successful as the previous 26. For that, the numerical analysis community owes its gratitude to the University of Strathclyde’s Numerical Analysis and Scientific Computing Group and, in particular, Philip Knight, John Mackenzie and Alison Ramage.

sdr
The conference dinner took place in one of Glasgow’s historic buildings: the Glasgow Trades Hall.