19 datasets found

A large-scale dataset of No Press Diplomacy games, containing more than 150,000 human games.

The dataset is used to evaluate agents for the Hanabi card game, a cooperative card game with imperfect information.

The dataset used in the paper is a collection of 6 publicly available datasets from the UCI Machine Learning repository.

The MineRL-v0 dataset contains human demonstration data for tasks in Minecraft.

The MineRL BASALT competition dataset contains human demonstration data for four tasks in Minecraft.

The dataset used in this work is the Breakout, SeaQuest, Space Invaders, and Qbert environments.

The Arcade Learning Environment (ALE) is a lasting and indispensable element of the RL researcher’s toolbox. It is also the focus of our work. Since its inception, hundreds of...

Pommerman is a 2-vs-2 game environment introduced in 2018, and is a modified version of the popular game 'Bomberman'. It is a good testbed for RL, because the environment...

The Arcade Learning Environment

The dataset used in this paper is the Car Racing game dataset, which consists of pixel frames of a car racing game.

The dataset used in this paper is the Atari 2600 games dataset, which consists of 50 Atari 2600 games.

The Atari 2600 domain dataset, used for training and testing reinforcement learning algorithms.

The dataset used in the paper is the Atari 2600 game dataset, which consists of 4 consecutive 80x80 gray-scale game frames as the input to the network.

The Atari 57 benchmark is a collection of 57 Atari games, each with its own set of states and actions.

The dataset used in the Generative Adversarial Exploration for Reinforcement Learning paper.

The dataset used in the paper is the Atari 2600 dataset, which consists of 49 games. The dataset is used to test the Successor Uncertainties algorithm.

The dataset contains data from human-level control through deep reinforcement learning.

The Million Base dataset is a large dataset of chess games, containing 2.5 million games.

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used several continuous control environments from the OpenAI Gym.