Skip to content

Examples

Four self-contained ways into ppsds. Each lives in its own directory under examples/. Every number these pages mention is printed by the script that produces it, so you can run the script and compare the output instead of trusting the text.

Example The story Who it is for Time
understand_ppsds_by_pmf_betting Place bets on where a system goes next, then find out what they were worth Anyone; no prerequisites 30 seconds, no install
apply_drpp_to_your_own_env Put the predictor on a system you already step — and find out when it pays and when it charges you You have a Gymnasium environment and intend to use the result a few seconds per script
understand_drpp_on_simulated_ambiguity Declare an ambiguity set, grow a system out of it, and watch three predictors face two meanings of “worst” You want the central semantics of DRPP, not just the API seconds
understand_predictability The ceiling no predictor can beat, and the paper's figure You want the number every predictor is measured against about a second

The first one needs no installation at all. The other three run on plain pip install ppsds; each directory carries its own environment.yml if you would rather create a conda environment for it.

Several examples have a heavier companion under Experiments: the same idea over thousands of trajectories, with a cache and a command line. The example is what you read to understand it; the experiment is what you run to produce numbers.

Next