pdg-control
Has become a variety of different directions:
Policy Costs
- Considering different norms. L1 L2 Assymetric fixed
- When can policy costs pay for themselves? notes
Additional Noise
- Starts with replicating Sethi et al. (2005).
- Consider the performance of various non-optimal solutions, particularly regarding imperfect implementation
- These are losely collected in Robustness examples, needs some evaluation of the most interesting.
Approaches to parameter, structural, and state uncertainty
- (Morph of) Training problem II from the working group.
- Active adaptive management using grid, model choice example.
- Add: Myopic Bayes passive adaptive management for model choice
- Compare to: stationary (no learning) solution averaging over models
Warning signals
- Accepted!
- Copy up on arxiv.
- Streamline code and examples. Began updating functions to work in a more generic manner that is better for connecting in pipelines and swapping between indicators. Move examples over from wiki to
inst/examples. - What to do with older repo
warningsignalsvs newer codebaseearlywarning? - Provide a generic early warning signals package? Perhaps Vasilis would be interested in helping?
Fallacy
- Write introduction with more basic background on warning signals
- Touch up discussion, edits.
- Updated runs from codebase. Heavy computation makes quick replication difficult
- draft sent to Alan