ml reinforcment tearning rl - ghdrako/doc_snipets GitHub Wiki

rl

Classical Algorithms

Classical algorithms are those algorithms that do not require deep neural networks for learning. They employ foundational approaches of RL to learn ideal behaviour through interaction with the environment via trial and error. Below is the list of a few classical algorithms:

  • Dynamic programming
  • Monte Carlo (MC) methods
  • Temporal difference (TD) learning
  • Q-learning
  • SARSA (state-action-reward-state-action)
  • Policy gradient methods