Title :
Learning, optimizing, and distributed decision making based on experience
Author_Institution :
Harvard Univ., Cambridge, MA, USA
Abstract :
We present a short and simplified "derivation" and discussion of perturbation analysis (PA), Markov decision problems (MDP), and reinforcement learning (RL) based on the sample path approach.
Keywords :
Markov processes; distributed decision making; learning (artificial intelligence); optimisation; perturbation techniques; Markov decision problems; distributed decision making; experience; optimization; perturbation analysis; reinforcement learning; sample path method; Contracts; Convergence; Costs; Distributed decision making; Dynamic programming; History; Learning; Stability analysis; State-space methods; Steady-state;
Conference_Titel :
Decision and Control, 2003. Proceedings. 42nd IEEE Conference on
Print_ISBN :
0-7803-7924-1
DOI :
10.1109/CDC.2003.1272352
