Should we have a common abstraction language?

Common abstraction language

  • everything should come down to SLD
  • a common vocabulary and grammar
  • instruction about how to arrange

Some constraints are different from others

  • But constraints are important in optimisation

Model builder to reflectometry calculation engine

  • Need to consider the constraints in between this (including mathematical descriptions)

A declaritive language that is translated to something that is optimised (with feedback to the GUI)

  • This will be a pain but useful for historical code and futureproofing
  • It would also improve the reusability of the different models.

It is important to consider atomistic/coarse-grained simulations as input.

What novel technologies should we be investigating?

Machine learning is sexy; but is it worthwhile?

  • Rapid prototyping of analysis
  • Check sense of models

Analysis is more powerful with data science tools (Bayesian inference, information theory)

Need to engage users with these methods

  • Particularly good for non-unique answers
  • Resources for education

Machine learning is the start of the process, not the end.

  • Currently a bit of a black box
  • Useful for the automation of “easy” tasks
  • Might be useful for assessing analysis feasibility at proposal stage