rl-practicum-I054 0.1

A simple package providing common Reinforcement Learning utility functions

RL Utils

A simple and useful Python package providing common Reinforcement Learning utility functions.

Installation

pip install rl-utils

Features

This package provides essential utility functions for implementing Reinforcement Learning algorithms:

• Epsilon-Greedy Action Selection: Classic exploration-exploitation strategy
• Q-Learning Update: Implementation of the Q-learning algorithm update rule
• Softmax Action Selection: Boltzmann exploration strategy
• Q-Table Initialization: Helper function to initialize Q-tables

Usage

Epsilon-Greedy Action Selection

from rl_utils import epsilon_greedy_action
import numpy as np

q_values = [0.5, 0.3, 0.8, 0.2]
action = epsilon_greedy_action(q_values, epsilon=0.1, num_actions=4)

Q-Learning Update

from rl_utils import q_learning_update, initialize_q_table
import numpy as np

# Initialize Q-table
q_table = initialize_q_table(num_states=10, num_actions=4)

# Update Q-value
q_table, new_q = q_learning_update(
    q_table=q_table,
    state=0,
    action=1,
    reward=10.0,
    next_state=2,
    alpha=0.1,
    gamma=0.9,
    num_actions=4
)

Softmax Action Selection

from rl_utils import softmax_action_selection

q_values = [0.5, 0.3, 0.8, 0.2]
action = softmax_action_selection(q_values, temperature=1.0)

Requirements

• numpy >= 1.11.1

License

MIT License

Author

Created for RL Practicum course.