OBP Package v0.5.0 released

[Open Bandit Project]

Dear all,

We have released the OBP package version 0.5.0: https://github.com/st-tech/zr-obp/releases/tag/0.5.0 

The changes are summarized below:

Major updates

• Add OPE/OPE with Continuous Actions
  • SyntheticContinuousBanditDataset (#112 )
  • ContinuousOPEEstimators [1] (#113 )
  • ContinuousNNPolicyLearner [2] (#114 )
• Add Weight clipping to IPW and DR (#115 )
• Add Automatic Hyperparameter Tuning of OPE estimators [3] (#116, #131 )
• Add some arguments to the SyntheticBanditDataset class to generate more flexible synthetic data (#123 )
• Add a subsample option to the OpenBanditDataset class (#118 )
• Modify an input type of off_policy_objective argument and Add some hyperparameters to NNPolicyLearner (#132)

Minor updates

• Fix README (#119 )
• Fix Scalar value checking (#122 )
• Add ValueError to OffPolicyEvaluation class (#125 )
• Fix Error messages (#126 )
• Add Some Errors (#125, #129 )
• Update Quickstart examples (#127 )

Cautions

• the hyperparameter name of obp.ope.SwitchDoublyRobust has changed to lambda_ from tau
• the type of argument off_policy_objective of obp.policy.NNPolicyLearner has changed to str from callable

References

• Nathan Kallus and Angela Zhou. Policy Evaluation and Optimization with Continuous Treatments, AISTATS2018.
• Nathan Kallus and Masatoshi Uehara. "Doubly Robust Off-Policy Value and Gradient Estimation for Deterministic Policies", NeurIPS2020.
• Yi Su, Maria Dimakopoulou, Akshay Krishnamurthy, and Miroslav Dudik. "Doubly Robust Off-Policy Evaluation with Shrinkage.", ICML2020.

Please update your package accordingly. We continue to improve and expand the software; stay tuned!

Best Regards,
Open Bandit Project Team