State cloning with the GraphLearner

This paper introduces the Cloned GraphLearner, a neuromorphic sequence generation model that mitigates state aliasing in high-order Markov chains through a lightweight, iterative state cloning procedure. Starting from the original GraphLearner, which stores variable length histories in a Graph Structured Bloom Filter, the algorithm successively creates layers of cloned states whose identity and inter-clone edges index increasingly long context windows while an oblivion rule bounds growth. When trained with action-observation sequences the resulting Cloned GraphLearner acts as a topographic schema with individual clones firing in context specific patterns that resemble hippocampal place cell activity.

Metadaten
Author:	Timothy Harrison, Herwig Unger
URN:	urn:nbn:de:bvb:384-opus4-1252576
Frontdoor URL	https://opus.bibliothek.uni-augsburg.de/opus4/125257
Parent Title (English):	3rd Workshop on Machine Learning in Networking (MaLeNe), co-located with the 6th International Conference on Networked Systems (NetSys 2025), Ilmenau, Germany, September 1, 2025: proceedings
Publisher:	Universität Augsburg
Place of publication:	Augsburg
Editor:	Michael SeufertORCiD GND, Andreas Blenk, Björn Richerzhagen
Type:	Conference Proceeding
Language:	English
Date of Publication (online):	2025/09/15
Year of first Publication:	2025
Publishing Institution:	Universität Augsburg
Release Date:	2025/09/16
Institutes:	Fakultät für Angewandte Informatik
	Fakultät für Angewandte Informatik / Institut für Informatik
	Fakultät für Angewandte Informatik / Institut für Informatik / Lehrstuhl für Vernetzte Systeme und Kommunikationsnetze
Dewey Decimal Classification:	0 Informatik, Informationswissenschaft, allgemeine Werke / 00 Informatik, Wissen, Systeme / 004 Datenverarbeitung; Informatik
Licence (German):	CC-BY 4.0: Creative Commons: Namensnennung (mit Print on Demand)

Open Access