As predicted in my my preview post, this conference, held on the Boston waterfront, proved to be SIAM’s largest ever, with 1378 attendees. Over 1000 presentations were given in up to 20 parallel minisymposia at a time, but this did mean that there was at least one talk (and usually several) of interest to me in almost every time slot.
One thing I learned from the conference is how widely Python is being used in computational science, especially for solving real world problems involving large amounts of data. This is partly due to its ability to act as the glue between codes written in other languages and web applications. The IPython environment, with its notebook interface, was featured in a number of talks, in some of which the slides were displayed using the notebook.
The following highly selective photos will give a flavour of the conference.
The conference venue. Note the residual snow, which fortunately did not fall in any serious amounts during the conference.
The poster session of about 65 posters was preceded by a poster blitz (1 minute presentations) and was accompanied by an excellent dessert. This photo shows Edvin Deadman (University of Manchester and NAG Ltd.) discussing his poster on Matrix Functions and the NAG Library with Cleve Moler and Charlie Van Loan (authors of the classic Nineteen Dubious Ways to Compute the Exponential of a Matrix paper). For some thoughts on poster sessions by one of the conference attendees see Please, no posters! by David Gleich.
Josh Bloom’s (UC Berkeley) invited presentation Automated Astrophysics in the Big Data Era contained a fascinating mix of observational astronomy, machine learning, robotic telescopes, numerical linear algebra, and Python, with a focus on classifying stars.
It was interesting to see MapReduce being used to implement numerical algorithms, notably in the minisymposium Is MapReduce Good for Science and Simulation Data? organized by Paul Constantine (Stanford; standing) and David Gleich (Purdue; sitting, with pointer).
Here is the lunchtime panel Big Data Meets Big Models being videod. Highlights from this panel and some of the invited plenary talks will be available in due course on the SIAM Presents YouTube channel.
If you weren’t at the conference perhaps you can make it to the next one in two year’s time (date and location to be announced). In the meantime a good way to keep up with events is to join the SIAM Activity Group on Computational Science and Engineering, which organizes the conference.
Nick Higham 
> One thing I learned from the conference is how widely Python is being used
> in computational science
Yes, and your published algorithms are slowly making their way into scipy (the main scientific python library). In my ideal world you would write python instead of matlab supplemental code (or make your matlab code scipy-license-compatible) but your published papers are detailed enough that people can write the python code after reading the algorithm descriptions in your papers, which is the next best thing!
matrix square root:
https://github.com/scipy/scipy/pull/2532
frechet derivative of matrix exponential:
https://github.com/scipy/scipy/blob/master/scipy/linalg/_expm_frechet.py
action of matrix exponential:
https://github.com/scipy/scipy/blob/master/scipy/sparse/linalg/_expm_multiply.py
1-norm estimate:
https://github.com/scipy/scipy/blob/master/scipy/sparse/linalg/_onenormest.py