Skip to content

Research collaborations

The last column of Get started is only data: you have rollouts but no dynamics model, or you work on a system this library does not cover yet. L0–L3 are deliberately empty there. That column is future research, and a discussion is the right first contribution.

What is worth discussing

  • A system or a domain the conic-moment family cannot express. State-dependent or multiplicative noise, correlated noise, or a constraint that is not additive are all outside the current family. A small, honest description of the gap is more useful than a general offer.
  • A new family of ambiguity sets, and the DRPP it implies. The library is built for this: each family is a new module at the same level in four places. The developer guide is the end-to-end contract.
  • A new kind of probabilistic predictor. The predictor contract is the ProbabilisticPredictor protocol; a predictor that predicts a pmf you supply is already the second kind in the package.
  • A scoring rule or a predictability question that changes what "good" means. Start from the probabilistic predictability paper.

How to start

Open an issue at https://github.com/Xuzerken/ppsds/issues describing the system, the data you have, and the question you want the prediction to answer. If you already have a derivation or a prototype, say so; the useful next step is usually a contract, not a full implementation.

The published surface deliberately does not describe unpublished methods. If your work depends on one, say that you have it and discuss the interface, not its internals.

Next