Reinforcement Learning

Environments, agents, and training algorithms (DQN, REINFORCE, A2C, PPO, SAC) built on the Deep Learning Modules and Layers tensor/autograd core. More...

Files
file	agent.hpp
	Abstract RL agent interface – the inference-time policy contract, act() only.
file	cartpole_env.hpp
	CartPole-v1 environment – classic cart-pole balancing physics, discrete actions.
file	categorical_policy_agent.hpp
	Stochastic categorical (discrete-action) policy over a logit-producing Module.
file	continuous_cartpole_env.hpp
	Continuous-action variant of CartPoleEnv – identical physics, force is a fraction.
file	dqn_agent.hpp
	Epsilon-greedy DQN policy over an arbitrary Q-network Module.
file	dqn_loss.hpp
	DQN's masked-MSE Bellman loss – squared error on the taken action only.
file	dqn_target.hpp
	DQN Bellman target computation (vanilla + Double DQN) and target-network hard sync.
file	environment.hpp
	Abstract RL environment interface (gymnasium-shaped reset/step) + StepResult.
file	gae.hpp
	Generalized Advantage Estimation (Schulman et al. 2016) over a stored rollout.
file	policy_gradient_loss.hpp
	REINFORCE's return-weighted negative log-likelihood loss over a batched rollout.
file	polyak_update.hpp
	Soft (Polyak / exponential-moving-average) target-network update, as used by SAC.
file	ppo_clipped_loss.hpp
	PPO's clipped surrogate objective (Schulman et al. 2017) over a batched rollout.
file	replay_buffer.hpp
	Off-policy experience replay: fixed-capacity circular transition store + sampling.
file	rollout_buffer.hpp
	On-policy trajectory storage: fill-once fixed-length rollout + discounted returns.
file	tanh_gaussian_policy.hpp
	SAC's reparameterized, tanh-squashed Gaussian policy sampling (action + log-prob).

Detailed Description

Environments, agents, and training algorithms (DQN, REINFORCE, A2C, PPO, SAC) built on the Deep Learning Modules and Layers tensor/autograd core.