GaussianBandit#

statrl.settings.bandits.stochastic.anytime.envs.parametric.GaussianBandit(means, vars, name='MAB-Gaussian')[source]#

Build a Gaussian bandit from vectors of means and variances.

Parameters:

  • means (array-like of float) – Mean reward of each arm.

  • vars (array-like of float) – Variance of each arm; must have the same length as means.

  • name (str, default='MAB-Gaussian') – Prefix of the instance name. Only the means are appended to it, so two instances differing solely in their variances share a name — pass a distinct name when comparing those.

Returns:

A bandit with unbounded rewards. Pair it with IMED using klGauss().

Return type:

StochasticBanditEnv