Policies.Experimentals package¶

Policies.Experimentals module : contains experimental or unfinished (single-player) bandits algorithms:

Index based UCB algorithms: UCBlog10, UCBwrong, UCBlog10alpha, UCBcython, UCBjulia,
Based on Kullback-Leibler divergence: klUCBlog10, klUCBloglog10,
Empirical KL-UCB algorithm: KLempUCB (does not work with the C optimized version of kullback),
An experimental policy, using Unsupervised Learning: UnsupervisedLearning,
An experimental policy, using Black-box optimization: BlackBoxOpt,
Bayesian algorithms: ThompsonRobust,
New! The UCBoost (Upper Confidence bounds with Boosting) policies, first with no boosting, in module UCBoost_faster: UCBoost_faster.UCB_sq, UCBoost_faster.UCB_bq, UCBoost_faster.UCB_h, UCBoost_faster.UCB_lb, UCBoost_faster.UCB_t, and then the ones with non-adaptive boosting: UCBoost_faster.UCBoost_bq_h_lb, UCBoost_faster.UCBoost_bq_h_lb_t, UCBoost_faster.UCBoost_bq_h_lb_t_sq, UCBoost_faster.UCBoost, and finally the epsilon-approximation boosting with UCBoost_faster.UCBoostEpsilon. These versions use Cython for some functions.
New! The UCBoost (Upper Confidence bounds with Boosting) policies, first with no boosting, in module UCBoost_cython: UCBoost_cython.UCB_sq, UCBoost_cython.UCB_bq, UCBoost_cython.UCB_h, UCBoost_cython.UCB_lb, UCBoost_cython.UCB_t, and then the ones with non-adaptive boosting: UCBoost_cython.UCBoost_bq_h_lb, UCBoost_cython.UCBoost_bq_h_lb_t, UCBoost_cython.UCBoost_bq_h_lb_t_sq, UCBoost_cython.UCBoost, and finally the epsilon-approximation boosting with UCBoost_cython.UCBoostEpsilon. These versions use Cython for the whole code.

Policies.Experimentals package¶

Submodules¶