List of benchmarks / toy problems for possible inclusion

[ben18785]

Update by Michael: I've repurposed this ticket to collect various toy problems / benchmarks we might like to add at some point. Whenever someone decides a certain benchmark is necessary - and that they're going to implement it - we can make a separate ticket for that!

• Hass et al. (2018 pre-print) - Reaction-diffusion problems, Optimisation focus
  • 20 biochemical reaction networks, all taken from literature & with real data.
  • Note in discussion that log transform is good because convexity changes
• Villaverde et al. "BioPredyn" (2015) - All reaction-diffusion problems, Optimisation focus
  • Problem B1: genome-wide kinetic model of S. cerevisiae
  • Problem B2: dynamic model of the Central Carbon Metabolism of E. coli
  • Problem B3: enzymatic and transcriptional regulation of the Central Carbon Metabolism of E. coli
  • Problem B4: metabolic model of Chinese Hamster Ovary (CHO) cells
  • Problem B5: signal transduction logic model
  • Problem B6: the gap gene network of the vinegar fly, Drosophila melanogaster
• Ballnus et al (2017). - All reaction-diffusion problems, MCMC focus
  • Very nice plots of the posterior (or a representative cut) for each problem
  • M1 mRNA transfection
  • M2 Bistable switch
  • M3 Saturated growth
  • M4 Biochemical system with Hopf bifurcation
  • M5 Driven Van der Pol Oscillator
  • M6 Lorenz Attractor
• Feroz & Skilling (2013) - Theoretical focus
  • Rastrigin function,
  • Himmelblau's function,
  • multidimensional eggbox,
  • Bayesian object detection case study.
• Wikipedia's list of optimisation test functions

[MichaelClerx]

One the one hand, I'm all for including these. But on the other hand, every new toy problem means another row/column in our performance testing matrix, plus a lot more code to document and maintain. We have to be on the look-out for bloat!

[ben18785]

Nice list of functions commonly used in mathematical optimisation.

Agreed @MichaelClerx, just listing stuff which we can decide about later.

[ben18785]

For MCMC, not sure that Himmelblau will be that useful although may be for optimisation.

[MichaelClerx]

@chonlei found that Villaverde et al. have done another optimisation benchmark paper https://academic.oup.com/bioinformatics/article/35/5/830/5078478