Experiments¶
An experiment is the rigorous half of an example: the same question, at a scale where the answer is a number rather than an impression. Every experiment here is reproducible, states its runtime, and is the code behind a figure or a table in a paper.
| Experiment | Deepens | Produces | Runtime | Paper |
|---|---|---|---|---|
| drpp | what “worst case” means | cached pickles, error bars, covariance ellipses, seven models | minutes | conic DRPP paper |
| pmf_betting | the betting game | scored hands and exported runs, against the ceiling | seconds | uses the predictability bound |
| predictability | the ceiling | the three-panel figure over 100,000 trajectories | seconds | probabilistic predictability paper |
Each experiment directory carries its own environment.yml, so you can create exactly what one
experiment needs. The command lines below assume the repository root as the working directory, with
ppsds installed editable (python -m pip install -e ../.. from the experiment directory).
Next¶
- The examples these deepen: Examples
- The papers the figures belong to: Research
- The commands, once: Getting Started