Behavior Policies¶
Behavior policy estimation utilities.
BehaviorPolicyFit
dataclass
¶
Container for estimated behavior propensities and diagnostics.
apply(dataset)
¶
Return a dataset with estimated propensities attached.
behavior_diagnostics(action_probs, actions, propensities, *, clip_min=0.001, num_bins=10)
¶
Aggregate diagnostics for behavior policy estimation.
fit_behavior_policy(dataset, *, method='logit', model=None, model_kwargs=None, clip_min=0.001, seed=0, store_action_probs=False)
¶
Estimate behavior policy propensities from logged data.