Support2 Reinforcement Learning Benchmark Dataset

Support2 is a tabular reinforcement learning benchmark dataset hosted on OpenML. The dataset is categorized for reinforcement learning tasks, indicating it contains state, action, and reward information suitable for training and evaluating RL agents. Details on its creator, size, and specific temporal context are not provided in the available metadata.

Use Cases

• Benchmarking RL agent performance on state-action-reward transitions from the Support2 environment.
• Training policy networks using state features and reward signals to learn optimal action sequences.
• Comparing exploration strategies by analyzing agent trajectories and episode returns recorded in the data.
• Developing offline RL algorithms using logged experience data containing states, actions, and next states.

Strengths

• Part of the OpenML benchmark ecosystem, ensuring standardized evaluation.
• Categorized specifically for reinforcement learning, providing relevant structure for RL tasks.

Limitations

• Unknown row count and feature dimensions limit assessment of statistical power.
• Lack of column descriptions obscures the specific state and action spaces.

Provenance

Source

OpenML platform