Algorithms¶
Currently, EvoRL supports various training pipelines (workflows):
Reinforcement Learning (RL) Algorithms
Evolutionary Computation (EC) Algorithms, specific for policy search
Evolutionary Reinforcement Learning (EvoRL):
Evolution-guided Reinforcement Learning (ERL)
Population-based AutoRL
This document introduces these types of algorithms implemented in EvoRL. All algorithms are defined in evorl.algorithms.
RL Algorithms¶
Supported RL Algorithms:
Algorithm |
Workflow |
Policy Type |
Supported Action Space |
|---|---|---|---|
Random |
- |
Discrete & Continuous |
|
A2C |
Stochastic |
Discrete & Continuous |
|
PPO |
Stochastic |
Discrete & Continuous |
|
IMPALA |
Stochastic |
Discrete & Continuous |
|
DQN |
Value-based |
Discrete |
|
DDPG |
Deterministic |
Continuous |
|
TD3 |
Deterministic |
Continuous |
|
SAC |
Stochastic |
Discrete & Continuous |
|
TD7 |
Deterministic |
Continuous |
EC Algorithms¶
EC Algorithms are defines in the subpackage evorl.algorithms.ec.
Workflows for Single objective EC are derived from ECWorkflowTemplate.
Algorithm |
Workflow |
Policy Type |
Supported Action Space |
|---|---|---|---|
OpenES |
Deterministic |
Continuous |
|
VanillaES |
Deterministic |
Continuous |
|
ARS |
Deterministic |
Continuous |
|
CMA-ES |
Deterministic |
Continuous |
Workflows for Multi-objective EC are derived from MultiObjectiveECWorkflowTemplate. Currently, we provide NSGA-II with NSGA2Workflow for brax environments.
ERL Algorithms¶
The ERL algorithms are defined in the subpackage evorl.algorithms.erl. We provide ERL and CEM-RL and their variants.
Population-based AutoRL Algorithms¶
The Population-based AutoRL algorithms are defined in the subpackage evorl.algorithms.meta. We provide some general population-based training pipelines for RL hyperparameter tuning.