TruncatedGaussianBandit#

statrl.settings.bandits.stochastic.anytime.envs.parametric.TruncatedGaussianBandit(means, sigma=0.5, low=-1.0, high=1.0, name='MAB-TruncGaussian')[source]#

Build a bandit whose arms are Gaussians truncated to [low, high].

Parameters:

  • means (array-like of float) – Mean of each untruncated Gaussian. Each arm’s true mean is the truncated one, which differs whenever the bounds are not symmetric around means[a].

  • sigma (float, default=0.5) – Common standard deviation before truncation.

  • low (float, default=-1.0, 1.0) – Bounds of the reward support.

  • high (float, default=-1.0, 1.0) – Bounds of the reward support.

  • name (str, default='MAB-TruncGaussian') – Prefix of the instance name; sigma is appended to it.

Returns:

A bandit with rewards in [low, high].

Return type:

StochasticBanditEnv