Journal Title

Proceedings to The Twenty-Ninth International FLAIRS Conference

Publication Date

2016

Abstract

Episodic memory systems for artificially intelligent agents must cope with an ever-growing episodic memory store. This paper presents an approach for minimizing the size of the store by using specialized hash functions to convert each memory into a relatively short binary code. A set of desiderata for such hash functions are presented including locale sensitivity and reversibility. The paper then introduces multiple approaches for such functions and compares their effectiveness.

Subjects

Artificial intelligence; Episodic memory

Publication Information

Proceedings to The Twenty-Ninth International FLAIRS Conference, 2016, 116-121.

© 2016 Association for the Advancement of Artificial Intelligence

Archived version is the final published version.

Peer-Reviewed

No

Document Type

Conference Paper

Included in

Engineering Commons

Share

COinS