A case study of data-driven computational modeling in Epidemics: bringing the dirt to the classroom. Stefan Engblom from Uppsala University will present his research work and modeling principles for a national-scale computational model of Verotoxigenic Escherichia coli in the Swedish cattle population.

I will start by presenting our work towards a first principle
national-scale computational model of Verotoxigenic Escherichia coli
in the Swedish cattle population. Briefly, the model integrates
infectious dynamics as continuous-time Markov chains and ODEs with
data describing the detailed animal contact pattern consolidated on
the fly.

We have also developed a shared memory parallel solution algorithm
allowing highly resolved forward simulations covering 1.5M animals and
over 10 years of data to be performed in about half a minute on a
desktop multicore computer. These very features enable the solution of
inverse problems on national scales and our model has been
successfully fitted against observed data.

Leaving the dirt, I will continue the talk with some more personal
reflections. Under a holistic view on computations the very questions
sought to be answered through simulations must continuously be kept in
mind. We are taught that convergence requires well-posedness, but most
often the stability of determining model parameters is not discussed
at all.

I will conclude in the classroom by reviewing some recent work in
bringing a "Holistic Computing" perspective into our courses. The
ansatz consists of confronting the students with a single, but
arguably quite difficult project, notably centered around an inverse
problem. By breaking the project up into more manageable pieces, the
course material unfolds in a self-motivated way.

References:

[1] S. Engblom, Strong convergence for split-step methods in
    stochastic jump kinetics, in SIAM J. Numer. Anal. 53(6):2655--2676
    (2015).
[2] A. Chevallier, S. Engblom, Pathwise error bounds in Multiscale
    variable splitting methods for spatial stochastic kinetics,
    manuscript (2016), available at http://arxiv.org/abs/1607.00805.
[3] P. Bauer, S. Engblom, S. Widgren, Fast event-based epidemiological
    simulations on national scales, in Int. J. High
    Perf. Comput. Appl. 30(4):438--453 (2016).
[4] S. Widgren, S. Engblom, P. Bauer, J. Frössling, U. Emanuelson, and
    A. Lindberg: Data-driven network modelling of disease transmission
    using complete population movement data: spread of VTEC O157 in
    Swedish cattle, in Veterinary Res. 47:81(1):1--17, (2016).
[5] S. Widgren, P. Bauer, and S. Engblom: SimInf: An R package for
    data-driven stochastic disease spread simulations, manuscript
    (2016), available at http://arxiv.org/abs/1605.01421.
    See www.siminf.org.