認知の問題への力学系に基づくロボット構成論的アプローチ(一般)
スポンサーリンク
概要
- 論文の詳細を見る
The current paper reviews our ideas of dynamical systems approach to study the problems of symbol grounding and compositionality in embodied cognitive systems. Our scheme of using a dynamic neural network model called RNNPB is introduced in which its applications to organization of behavior primitives, segmentation and chunking in sensory-motor flow and co-learning of simple language and behavior are examined. These robotics experiments demonstrate that the RNNPB could organize certain combinatorial mechanism, as an alternative of a symbol system, of which interface to physical bodies and environments is much more flexible and adaptive than the ones by the conventional symbol systems.
- 社団法人電子情報通信学会の論文
- 2004-10-11
著者
関連論文
- Mixture of RNN expertsによるルールダイナミクスの学習(機械学習によるバイオデータマインニング,一般)
- 認知の問題への力学系に基づくロボット構成論的アプローチ(一般)
- リカレントネットワークのアトラクタと有限オートマトンの学習