st-tech / zr-obp

Open Bandit Pipeline: a python library for bandit algorithms and off-policy evaluation

st-tech/zr-obp Issues

Training on test set?
Updated 3 months ago
Wrong parameter set in fig.add_subplot() in OffPolicyEvaluation.visualize_off_policy_estimates_of_multiple_policies
Updated 5 months ago2
Missing parentheses in z-score calculation?
Updated 6 months ago
Improperly escaped docstrings
Updated a year ago
Self Normalized Estimator _estimate_round_rewards is wrong?
Updated 2 years ago
New simulation functionality
Updated 2 years ago
propensity score estimate
Updated 2 years ago2
Some values are hard-coded in the code (!!)
Updated 2 years ago
context vector dimensions are different for data sampled with random policy and those sampled with Bernoulli TS
Closed 2 years ago4
"action_context" arrays in bandit feedbacks & item_context.csv (downloaded from https://research.zozo.com/data.html#OpenBanditDataset)
Closed 2 years ago2
Question on the first three elements of context vector
Closed 2 years ago2
contents of "item_context.csv" in OBD are different for random policy and Bernoulli TS policy
Closed 2 years ago2
"action_context" arrays in bandit feedbacks of different policies have different values
Closed 2 years ago2
Could I get to know what each dimension in item_context.csv or "action_context" array from bandit feedback are representing?
Closed 2 years ago2
How to use kernelized inverse probability weighting with OBD and OBP
Closed 2 years ago2
Loading my own data?
Closed 3 years ago3
Question: Does switch-dr/dr-os beat dr method in mse?
Closed 3 years ago1
Question: What is the source of the construction of evaluation and behavior policies?
Closed 3 years ago1
Question: Is there an example usage of estimator_tuning?
Closed 3 years ago1
Question: Can we select different regression models for different estimators?
Closed 3 years ago3
Switch DR current estimator has a typo
Closed 3 years ago5
Question/Feature: Is it possible to add the varying examples/n from the datasets?
Closed 3 years ago1
Question/Feature: Is it possible to add the 10 UCI datasets as well for evaluating these OPEs?
Closed 3 years ago1
alpha_ and lambda_ are not necessary for contextual linear bandit algorithms
Closed 3 years ago2
The type check frustrate the policy value calculation
Closed 3 years ago3
Can I use this tool in RL(MDP) environments besides the bandit situation?
Closed 3 years ago1
what's the meaning of position in OPE?
Closed 3 years ago1
Add new github actions
Updated 3 years ago1
np.newaxis is simpler than np.expand_dims
Closed 3 years ago
IPWLearner fit method does not allow position=None
Closed 3 years ago1
We cannot set integers to the hyperparameters of Switch-DR and DRos
Closed 3 years ago1
`obp.dataset.BaseBanditDataset` class does not exist
Closed 3 years ago3
RegressionModel should use bandit_feedback_test in the quickstart notebook for the synthetic data
Closed 3 years ago3
Replay method allows only deterministic policies
Closed 3 years ago2
Current Read the Docs is not the latest version
Updated 3 years ago
Parameter eval_size in MultiClassToBanditReduction.obtain_batch_bandit_feedback exists only in the docstring
Closed 3 years ago1
Citation format lacks uniformity
Closed 3 years ago1
More precise logics for n_actions in Dataset and Simulator
Updated 4 years ago
Any specific code formatter?
Closed 4 years ago3
Quickstart is not working without preparing obd directory
Closed 4 years ago2
typo in quickstart
Closed 4 years ago1
Question: LinTS takes a hours to run
Closed 4 years ago2