Glossario IA
Il dizionario completo dell'Intelligenza Artificiale
Asynchronous Advantage Actor-Critic (A3C)
Distributed architecture where multiple agents train in parallel on copies of the environment, sampling uncorrelated trajectories and accelerating convergence.
Soft Actor-Critic (SAC)
Off-policy algorithm that maximizes based on expected reward and policy entropy, promoting exploration and better robustness to hyperparameter tuning.
Deep Deterministic Policy Gradient (DDPG)
Off-policy algorithm for continuous action spaces combining DQN and Actor-Critic, using target networks and a deterministic policy.
Twin Delayed DDPG (TD3)
Improvement of DDPG using two critic networks to reduce overestimation bias and delayed actor updates to increase stability.
Munchausen-RL
Algorithm introducing a logarithmic entropy term in the Q update, inspired by Munchausen's algorithm, improving exploration and stability.