Concept策略梯度算法 / Williams算法
REINFORCE Algorithm
One of the earliest policy-gradient reinforcement learning algorithms, proposed by Williams in 1992. It estimates policy gradients using Monte Carlo sampling and updates policy parameters directly via gradient ascent. REINFORCE serves as a theoretical foundation for more advanced algorithms such as PPO.
Timeline (last 90 days)
Sep 11
LETHE continuously evolves through the interaction of a five-feature linear discriminator and a stochastic perturbation optimizer, which resembles a single-sample REINFORCE algorithm and can adjust parameters without gradient information.
Unverified50%