decisionbandit 0.1.1

A collection of exploration–exploitation strategies for reinforcement learning, including ε-greedy and related policies

epsilon-policies

A collection of exploration–exploitation strategies for reinforcement learning, including ε-greedy and related policies.

Features

• Exploration – Select random actions to discover new possibilities.
• Exploitation – Choose the best-known action based on current estimates.
• Fixed Exploration–Then–Exploitation – Explore for a fixed period, then fully exploit.
• ε-Greedy – Balance exploration and exploitation with a probability parameter.
• ε-Greedy with UCB – Enhance ε-greedy with Upper Confidence Bound for better action selection.

Installation

pip install decisionbandit


import decisionbandit as dcb

# Example: ε-greedy

reward_prob=[0.2,0.3,0.4]
action = dcb.epsilon_greedy(reward_prob,epsilon=0.1, n_arms=3,t_steps=100)
print("Selected action:", action)


MIT License