Crate rlevo 

Deep Reinforcement Learning with Evolutionary Optimization built on the Burn framework.

Modules

• core — foundational traits: Environment, State, Action, Reward, TensorConvertible
• envs — benchmark environments: classic control, gridworlds, Box2D physics, locomotion
• rl — deep RL algorithms: DQN, C51, QR-DQN, PPO, PPG, DDPG, TD3, SAC (and the replay buffer / experience / metrics modules they consume)
• evo — evolutionary algorithms: GA, ES, EP, DE, CGP with GPU kernels
• hybrid — combined evolutionary + RL strategies

Quick Start

[dependencies]
rlevo = "0.1"

use rlevo::prelude::*;

For specific items use the sub-module paths directly:

use rlevo::core::environment::Environment;
use rlevo::envs::classic::cartpole::CartPole;
use rlevo::rl::algorithms::dqn::dqn_agent::DqnAgent;

Re-exports

pub use rlevo_core as core;

pub use rlevo_environments as envs;

pub use rlevo_evolution as evo;

pub use rlevo_hybrid as hybrid;

pub use rlevo_reinforcement_learning as rl;

Modules

prelude

The most commonly used traits and types, importable with use rlevo::prelude::*.