Policy: Epsilon Greedy

EpsilonGreedyPolicy chooses an arm at random (explores) with probability epsilon, otherwise it greedily chooses (exploits) the arm with the highest estimated reward.

Usage

policy <- EpsilonGreedyPolicy(epsilon = 0.1)

Arguments

epsilon

numeric; value in the closed interval (0,1] indicating the probablilty with which arms are selected at random (explored). Otherwise, EpsilonGreedyPolicy chooses the best arm (exploits) with a probability of 1 - epsilon

name

character string specifying this policy. name is, among others, saved to the History log and displayed in summaries and plots.

Methods

new(epsilon = 0.1)

Generates a new EpsilonGreedyPolicy object. Arguments are defined in the Argument section above.

set_parameters()

each policy needs to assign the parameters it wants to keep track of to list self$theta_to_arms that has to be defined in set_parameters()'s body. The parameters defined here can later be accessed by arm index in the following way: theta[[index_of_arm]]$parameter_name

get_action(context)

here, a policy decides which arm to choose, based on the current values of its parameters and, potentially, the current context.

set_reward(reward, context)

in set_reward(reward, context), a policy updates its parameter values based on the reward received, and, potentially, the current context.

References

Gittins, J., Glazebrook, K., & Weber, R. (2011). Multi-armed bandit allocation indices. John Wiley & Sons. (Original work published 1989)

Sutton, R. S. (1996). Generalization in reinforcement learning: Successful examples using sparse coarse coding. In Advances in neural information processing systems (pp. 1038-1044).

Strehl, A., & Littman, M. (2004). Exploration via model based interval estimation. In International Conference on Machine Learning, number Icml.

Yue, Y., Broder, J., Kleinberg, R., & Joachims, T. (2012). The k-armed dueling bandits problem. Journal of Computer and System Sciences, 78(5), 1538-1556.

See also

Core contextual classes: Bandit, Policy, Simulator, Agent, History, Plot

Bandit subclass examples: BasicBernoulliBandit, ContextualLogitBandit, OfflineReplayEvaluatorBandit

Policy subclass examples: EpsilonGreedyPolicy, ContextualLinTSPolicy

Examples

horizon            <- 100L
simulations        <- 100L
weights          <- c(0.9, 0.1, 0.1)

policy             <- EpsilonGreedyPolicy$new(epsilon = 0.1)
bandit             <- BasicBernoulliBandit$new(weights = weights)
agent              <- Agent$new(policy, bandit)

history            <- Simulator$new(agent, horizon, simulations, do_parallel = FALSE)$run()
#> Simulation horizon: 100
#> Number of simulations: 100
#> Number of batches: 1
#> Starting main loop.
#> Finished main loop.
#> Completed simulation in 0:00:00.943
#> Computing statistics.

plot(history, type = "cumulative")

plot(history, type = "arms")