takuseno/nnabla-drl-collections

Deep reinforcement learning implementations written in NNabla

NNabla DRL Collections

Deep reinforcement learning implementations with NNabla.

install

$ pip install -r requirements.txt

If you use GPU, see here.

algorithms (discrete action-space)

• Deep Q-Network (DQN)
• Double DQN
• Dueling DQN
• NoisyNet-DQN
• Prioritized Experience Replay
• Categorical DQN
• Bootstrapped DQN
• Rainbow
• Quantile Regression DQN (QR DQN)
• Implicit Quantile Networks (IQN)
• Advantage Actor-Critic (A2C)
• Proximal Policy Optimization (PPO)

algorithms (continuous action-space)

• Deep Deterministic Policy Gradients (DDPG)
• Twin Delayed Deep Deterministic Policy Gradients (TD3)
• Soft Actor-Critic (SAC)
• SAC (learned temperature)
• Proximal Policy Optimization (PPO)
• Trust Region Policy Optimization (TRPO)
• Actor-Critic with Experience Replay (ACER)

blog posts

• Deep Q-Network Implementation with SONY’s NNabla