NS-Gym Algorithms

NS-Gym has several built-in benchmark algorithms for training and evaluating agents in non-stationary environments. These algorithms are implemented in the ns_gym.benchmark_algorithms module.

Attention

Detailed descriptions of included algorithms pages are currently under development. For now please take a look at the API documentation for ns_gym.benchmark_algorithms.

NS-Gym has implementations of the following algorithms:

• Monte Carlo Tree Search

• Proximal Policy Optimization

• Deep Q-Network

• AlphaZero

• Policy Augmented MCTS

• ADA-MCTS (Code included and runnable on NS-Gym envs though needs to be refactored for API compliance)

• RATS (Code included and runnable on NS-Gym envs but needs to be refactored for API compliance)

We also support integration with Stable-Baselines3, allowing users to leverage a wide range of RL algorithms in NS-Gym environments.