Dynamic concept model learns optimal policies

Author

Szepesvári, Cs

Author_Institution

Dept. of Math., Jozsef Attila Univ., Szeged, Hungary

Volume

3

fYear

1994

fDate

27 Jun-2 Jul 1994

Firstpage

1738

Abstract

Dynamic concept model (DCM) is a goal-oriented neural controller, that builds an internal representation of events and chains of events in the form of a directed graph and uses spreading activation for decision making. It is shown, that a special case of DCM is equivalent to reinforcement learning (RL) and is capable of learning the optimal policy in a probabilistic world. The memory and computational requirements of both DCM and RL are analyzed and a special algorithm is introduced, that ensures intentional behavior

Keywords

Markov processes; decision theory; directed graphs; learning (artificial intelligence); neurocontrollers; computational requirements; decision making; directed graph; dynamic concept model; goal-oriented neural controller; intentional behavior; internal representation; optimal policies; probabilistic world; spreading activation; Algorithm design and analysis; Artificial neural networks; Brain modeling; Control systems; Cost function; Decision making; Explosions; Learning; Mathematics; Problem-solving;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on

Conference_Location

Orlando, FL

Print_ISBN

0-7803-1901-X

Type

conf

DOI

10.1109/ICNN.1994.374418

Filename

374418