Learners¶
Each of reward_model, contextual_policy, and best_arm accepts
either a character string naming a built-in learner or a
user-supplied object implementing the required fit / predict
methods.
Built-in strings¶
| slot | strings | backend |
|---|---|---|
reward_model |
random_forest, linear, ridge, lasso |
scikit-learn |
contextual_policy |
policy_tree, policy_forest, linear (aliases dr_econml_2, dr_econml) |
econml |
best_arm |
ips, simple (aliases best_arm_erm, simple_best_arm) |
kpe |
User-supplied-object protocols¶
kpe.learners.base.RewardModel
¶
Bases: Protocol
kpe.learners.base.Policy
¶
Bases: Protocol
kpe.learners.base.BestArm
¶
Bases: Protocol