BayesUCB

Bayes-UCB bandit policy.

Bayes-UCB is a Bayesian algorithm for the multi-armed bandit problem. It uses the posterior distribution of the reward of each arm to compute an upper confidence bound (UCB) on the expected reward of each arm. The arm with the highest UCB is then pulled. The posterior distribution is updated after each pull. The algorithm is described in [^1].

Parameters

• reward_obj

  Default → None

  The reward object that is used to update the posterior distribution.

• burn_in

  Default → 0

  Number of initial observations per arm before using the posterior distribution.

• seed

  Type → int | None

  Default → None

  Random number generator seed for reproducibility.

Attributes

• ranking

  Return the list of arms in descending order of performance.

Examples

import gym
from river import bandit
from river import proba
from river import stats

env = gym.make(
    'river_bandits/CandyCaneContest-v0'
)
_ = env.reset(seed=42)
_ = env.action_space.seed(123)

policy = bandit.BayesUCB(seed=123)

metric = stats.Sum()
while True:
    action = policy.pull(range(env.action_space.n))
    observation, reward, terminated, truncated, info = env.step(action)
    policy = policy.update(action, reward)
    metric = metric.update(reward)
    if terminated or truncated:
        break

metric

Sum: 841.

Methods

compute_index

the p-th quantile of the beta distribution for the arm

Parameters

• arm_id

pull

Pull arm(s).

This method is a generator that yields the arm(s) that should be pulled. During the burn-in phase, all the arms that have not been pulled enough times are yielded. Once the burn-in phase is over, the policy is allowed to choose the arm(s) that should be pulled. If you only want to pull one arm at a time during the burn-in phase, simply call next(policy.pull(arms)).

Parameters

• arm_ids — 'list[ArmID]'

Returns

ArmID: A single arm.

update

Rewrite update function

Parameters

• arm_id
• reward_args
• reward_kwargs