紀要論文

好奇心に基づいたELMによる連想記憶モデルの強化

Enhancement of Associative Memory Model with Curiosity based ELM

jpn

24321192

https://doi.org/10.15002/00021519

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

13

1

6

2017-03-31

Version of Record

Conventionally, most of AI Systems learn and decide actions based on the situations. This is a kind of logical approach without consideration of psychological factors. This research takes account of human’s curiosity into memory model for intelligent things. As a learning model, it uses a network called Hopfield Network as a representation of associative memory. The associative memory is a memory system which learns and keeps associative concepts and relations between objects and concepts. This research is mainly focused on curiosity based Extreme Learning Machine (ELM) for enhancement of associative memory model. Based on the psychologist Berlyne's theory, the four elements, novelty uncertainly, conflict, surprise are indicators of one’s curiosity. The associative memory learning or updating strategy are determined by calculating quadruple curiosity vector values in Curiosity Extreme Learning Machine (C-ELM) so as to strengthen learning and memorizing of associative knowledge represented in Hopfield networks. A set of zoo data is applied to two designated scenarios of showing a child’s curiosities to animals in experiments and the experimental results of associative knowledge networks are reported with the remarks from various aspects referring to the C-ELM and its implementation.

Article

AA12746425