pockerman/cubeai

Library for reinforcement learning with c++

cubeAI

CubeAI is a C++ library containing implementations of various reinforcement learning, filtering and planning algorithms. The library documentation (under development) can be found here CubeAI The Python flavor of the library can be found at PyCubeAI.

The following is an indicative list of examples. More tutorials can be found at CubeAI.

Examples

Reinforcement learning

• Example 0: DummyAgent on MountainCar-v0
• Example 1: Toy Markov chain
• Example 2: Multi-armed bandit with epsilon-greedy policy
• Example 3: Multi-armed bandit with soft-max policy
• Example 4: Advertisement placement
• Example 6: Iterative policy evaluation on FrozenLake-v0
• Example 7: Policy iteration on FrozenLake-v0
• Example 8: Value iteration on FrozenLake-v0
• Example 9: SARSA on CliffWalking-v0
• Example 10: Q-learning on CliffWalking-v0
• Example 14: Expected SARSA on CliffWalking-v0
• Example 15: Approximate Monte Carlo on MountainCar-v0
• Example 16: Monte Carlo tree search on Taxi-v3
• Example 17: A* search on a road network from Open Street Map data
• Example 18: Double Q-learning on CartPole-v0

Path planning

• Example 19: Path planning with rapidly-exploring random trees (TODO)
• Example 20: Path planning with dynamic windows (TODO)

PyTorch based examples

• Example 11: Simple linear regression
• Example 12: DQN on CartPole-v0
• Example 13: Reinforce algorithm on CartPole-v0
• Example 21: Simple logistic regression

Installation

Installation instructions and dependencies can be found here.

Documentation dependencies

There are extra dependencies if you want to generate the documentation. Namely,

• Doxygen
• Sphinx
• sphinx_rtd_theme
• breathe
• m2r2