Other Estimators

These estimators are available for advanced users, checks, and method development. They sit outside the core focus, either because they overlap a core method, carry caveated public status, or are narrower in scope. The grouping follows the sources of complexity in Choosing and Comparing Estimators.

Read this page after the core menu. These methods are useful for specific computational, behavioral, or diagnostic needs, but each carries a narrower public evidence or interpretation boundary.

Estimator	Source of complexity it answers	Use
NNES	Large state space	Neural continuation value with finite reward parameters.
MPEC	Large state space	Constrained-optimization form of the DDC likelihood.
UFXP	Large state space	Structural estimates without nesting a fixed point.
RHIP	Bounded or finite-horizon planning	Horizon-parameterised entropy IRL for route choice.
f-IRL	Reward recovery via state-marginal matching	f-divergence state-marginal method.
IQ-Learn	Imitation and inverse soft-Q	Inverse soft-Q learning diagnostics.

Research baselines under econirl.contrib: Max Margin Planning, GCL, GAIL, Deep MaxEnt IRL, Bayesian IRL.

• NNES
• MPEC
• UFXP
• RHIP
• f-IRL
• IQ-Learn