genriesz: Generalized Riesz Regression
genriesz is a Python package for Automatic Debiased Machine Learning (ADML) with Generalized Riesz Regression (GRR) under Bregman divergences, with automatic regressor balancing.
The code is available on https://github.com/MasaKat0/genriesz.
The central user-facing API is:
genriesz.grr_functional()for a generic linear functional,genriesz.grr_ate()for the ATE,genriesz.grr_att()for the ATT,genriesz.grr_did()for panel DID (implemented as ATT on ΔY),genriesz.grr_ame()for average marginal effects.
The package also provides:
genriesz.fit_density_ratio()for density ratio estimation (covariate shift) via generalized Bregman divergence minimization,nearest-neighbor matching as LSIF/Riesz regression (
genriesz.nn_matching_inverse_propensity_weights()and local-polynomial extensions),optional ScoreMatchingRiesz primitives in
genriesz.scorematchingriesz(requires PyTorch).