Crate asteria 

Asteria — a pure-Rust reinforcement learning and neural network library.

Three top-level modules cover the full stack:

• clab: tensor engine, activation/loss functions, weight initializers, and noise generators.
• core: graph-based neural networks, dense layers, parameter management, optimizers (Adam, RAdam, ADOPT, SGD), and LR schedulers (StepDecay, CosineAnnealing, LinearDecay, WarmupCosine, ReduceLROnPlateau).
• rl: environment trait, exploration strategies, and RL algorithms (Q-Learning, SARSA, DQN, AC, A2C, A3C, CACLA, DDPG, PPO, QAC, TD, ForwardModel, Metacritic).

Modules

clab

Tensor engine and mathematical utilities.

core

Graph-based neural network construction, parameter management, optimizers, and LR schedulers.

rl

Reinforcement learning algorithms, environments, and exploration strategies.