memorax.algorithms

Reinforcement learning algorithms for training agents.

PPO

PPO - Proximal Policy Optimization for discrete and continuous action spaces.

PPOConfig - Configuration dataclass for PPO.

PPOState - Training state for PPO.

MAPPO

MAPPO - Multi-Agent PPO for multi-agent environments.

MAPPOConfig - Configuration dataclass for MAPPO.

MAPPOState - Training state for MAPPO.

DQN

DQN - Deep Q-Network with double Q-learning.

DQNConfig - Configuration dataclass for DQN.

DQNState - Training state for DQN.

R2D2

R2D2 - Recurrent Experience Replay in Distributed RL.

R2D2Config - Configuration dataclass for R2D2.

R2D2State - Training state for R2D2.

SAC

SAC - Soft Actor-Critic for continuous control.

SACConfig - Configuration dataclass for SAC.

SACState - Training state for SAC.

PQN

PQN - Parallelised Q-Network (on-policy Q-learning).

PQNConfig - Configuration dataclass for PQN.

PQNState - Training state for PQN.

StreamAC

StreamAC - Actor-Critic with eligibility traces.

StreamACConfig - Configuration dataclass for StreamAC.

StreamACState - Training state for StreamAC.

GradientPPO

GradientPPO - PPO with gradient eligibility traces.

GradientPPOConfig - Configuration dataclass for GradientPPO.

GradientPPOState - Training state for GradientPPO.