非マルコフ的タスクへの適用を考慮した遺伝的プログラミングに関する実験的考察

金本 陽祐; 三好 雄一; 小野 典彦; 最上 義夫

doi:10.11509/sci.SCI08.0.104.0

The 52nd Annual Conference of the Institute of Systems, Control and Information Engineers

Session ID : 3F3-6

DOI https://doi.org/10.11509/sci.SCI08.0.104.0

Conference information

Host: Institute of Systems, Control and Information Engineers

On Genetic Programming Considering Its Applicability to Non-Markovian Tasks

*Yousuke Kanemoto, Yuichi Miyoshi, Norihiko Ono, Yoshio Mogami

Author information

Keywords: genetic programming, non-Markovian tasks, autonomous agents, evolutionary computation, machine learning

CONFERENCE PROCEEDINGS FREE ACCESS

Details

Abstract

In conventional genetic programming (GP) schemes, it has not been attempted to design an autonomous agent with internal states, which makes difficult to apply those schemes to non-Markovian tasks where the sensory inputs are ambiguous. We have proposed a GP scheme that evolves an autonomous agent with internal states needed to appropriately perform non-Markovian tasks. The potentials and limitations of the proposed scheme are discussed through its application to several non-Markovian tasks.

Corresponding author

Register with J-STAGE for free!