Experiments

Read more in the user guide.

Entry point

massiveruns.runLargeMulticoreExperiment

Benchmark several agents on one environment and plot their regret.

Pipeline

The stages runLargeMulticoreExperiment() chains together. Call them directly only to reuse one stage on its own.

onerun.Interaction	Base class for the interaction loop of a setting.
parallelruns.multicoreRuns	Run one agent for many independent replicates, spread across CPU cores.
onerun.oneRunWithDump	Run one replicate and pickle its score series to disk.
analyzeruns.computeScoreDiffs	Turn per-replicate score dumps into regret statistics over time.
plotruns.plotScoreDiffs	Draw the regret figures and record final regrets in the logfile.

Persistence

utils.dump	Pickle a value to {root_folder}{filename}_{tag}.
utils.clear_auxiliaryfiles	Delete the intermediate per-replicate dumps of one environment.
utils.load	Read an environments.yaml registry.
utils.make	Instantiate an environment from a registry spec.