The transition matrix of states seems wrong when we set "prob_leave_bandit"

criteo-research / reco-gym

Code for reco-gym: A Reinforcement Learning Environment for the problem of Product Recommendation in Online Advertising

Apache License 2.0

465 stars 98 forks source link

Open yxliu-ntu opened 4 years ago

yxliu-ntu commented 4 years ago

Should the line 56th of the file "reco_env_v1.py" be [self.config.prob_bandit_to_organic, 0, self.config.prob_leave_bandit] ?

ihtiihti commented 4 years ago

Yes, please, fix it.