Overview

The player is presented with an alien mother ship, which continually deploys three smaller ships during play.[2] The mother ship and the smaller vessels shoot at a weapon the player is in command of, and the player’s aim is to eliminate the opposition while preventing the weapon from receiving enough fire to destroy it.[2] The player uses a joystick to operate the game, and only one player at a time can play.[1]

Description from Wikipedia

State of the Art

Human Starts

Result Method Type Score from
24404.6 ApeX DQN DQN Distributed Prioritized Experience Replay
14497.9 A3C LSTM PG Asynchronous Methods for Deep Learning
14491.7 RainbowDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
10950.6 DuelingPERDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
7748.5 PERDDQN (prop) DQN Prioritized Experience Replay
6548.9 PERDDQN (rank) DQN Prioritized Experience Replay
6060.8 DDQN DQN Deep Reinforcement Learning with Double Q-learning
5474.9 A3C FF (4 days) PG Asynchronous Methods for Deep Learning
5124.3 NoisyNetDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
5101.3 DistributionalDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
3994.8 DuelingDDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
3746.1 A3C FF (1 day) PG Asynchronous Methods for Deep Learning
3489.3 DQN2015 DQN Dueling Network Architectures for Deep Reinforcement Learning
3332.3 DQN2015 DQN Massively Parallel Methods for Deep Reinforcement Learning
3081.3 PERDQN (rank) DQN Prioritized Experience Replay
1195.85 GorilaDQN DQN Massively Parallel Methods for Deep Reinforcement Learning
628.9 Human Human Massively Parallel Methods for Deep Reinforcement Learning
166.9 Random Random Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Method Type Score from
24559.4 ApeX DQN DQN Distributed Prioritized Experience Replay
14198.5 RainbowDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
11477.0 DuelingPERDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
11231.0 NoisyNet-DuelingDQN DQN Noisy Networks for Exploration
10777.7 ACKTR PG Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
9011.6 DDQN+PopArt DQN Learning values across many orders of magnitude
8010.0 DuelingDQN DQN Noisy Networks for Exploration
7965.7 PERDDQN (prop) DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
7672.1 PER DQN Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
7672.1 PERDDQN (rank) DQN Dueling Network Architectures for Deep Reinforcement Learning
7203.0 C51 Misc A Distributional Perspective on Reinforcement Learning
5909.0 DistributionalDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
5510.0 NoisyNet-DQN DQN Noisy Networks for Exploration
5393.2 DDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
5198.6 NoisyNetDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
5022.9 DDQN DQN Deep Reinforcement Learning with Double Q-learning
4621.0 DuelingDDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
4280.4 DQN2015 DQN Dueling Network Architectures for Deep Reinforcement Learning
3595.0 DQN DQN Noisy Networks for Exploration
3359 DQN2015 DQN Human-level control through deep reinforcement learning
3060.0 NoisyNet-A3C PG Noisy Networks for Exploration
2879.0 A3C PG Noisy Networks for Exploration
1880.9 DuelingPERDDQN DQN Deep Q-Learning from Demonstrations
1755.7 DQfD Imitation Deep Q-Learning from Demonstrations
1496 Human Human Human-level control through deep reinforcement learning
1450.41 GorilaDQN DQN Massively Parallel Methods for Deep Reinforcement Learning
742.0 Human Human Dueling Network Architectures for Deep Reinforcement Learning
628 Linear Misc Human-level control through deep reinforcement learning
537 Contingency Misc Human-level control through deep reinforcement learning
222.4 Random Random Human-level control through deep reinforcement learning

Normal Starts

Result Method Type Score from
4971.9 PPO PG Proximal Policy Optimization Algorithms
4653.8 ACER PG Proximal Policy Optimization Algorithms
1562.9 A2C PG Proximal Policy Optimization Algorithms