Model variations

[cboettig]

Just starting to collect a list of model conditions we will likely want to explore:

• Temporal heterogeneity? What happens if the model changes

• Transfer learning / robustness of policy: how well does a policy trained in one environment do in another?
  (For comparison, the technical optimal control solution of constant escapement is quite robust to model details in some ways, i.e. optimal escapement level is K/2 for any symmetric model, but obviously this is quite sensitive to asymmetry, inaccurate K, or imperfect measurements

• Tipping point growth dynamics

• ....