KNOWLEDGE GRAPH CONSTRUCTION AND MACHINE LEARNING FOR IMPROVED THREAT IDENTIFICATION
Loading...
Authors
Kimball, Jessica
Subjects
knowledge graph construction
machine learning
dynamic graph neural networks
probabilistic inference
semantic extraction
Koopman operator
semi-supervised
large language models
retrieval-augmented generation
deep graph representation learning
Semi-supervised Predictive Autoencoder Representation using Koopman Learning Evolution
SPARKLE
artificial intelligence
AI
machine learning
dynamic graph neural networks
probabilistic inference
semantic extraction
Koopman operator
semi-supervised
large language models
retrieval-augmented generation
deep graph representation learning
Semi-supervised Predictive Autoencoder Representation using Koopman Learning Evolution
SPARKLE
artificial intelligence
AI
Advisors
Gallup, Shelley P.
Blais, Curtis L.
Bordetsky, Alex
Mun, Johnathan C.
Zhao, Ying
Date of Issue
2025-03
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
Knowledge reasoning and representation in artificial intelligence (AI) are pivotal for advancing predictive research in threat identification. The rapid increase in large-scale data has spurred the deployment of automated solutions, yet current machine learning interfaces still struggle to reliably predict anomalous behaviors—limiting their suitability for critical decision-making. To address this challenge, recent advances in graph neural network theory and modern Koopman theory for dynamical systems have enabled the development of deep graph representation learning techniques combined with knowledge graph construction. This approach enhances threat classification accuracy by learning graph embeddings that capture outlier threat scores. Iterative comparisons using graph similarity measures between the predicted generative graph and the ground truth further refine the predictions. Dimensionality reduction is achieved using Koopman on violent incident information from news articles. The proposed Semi-supervised Predictive Autoencoder Representation using Koopman Learning Evolution (SPARKLE) method offers a scalable, adaptive framework for evolving intelligence, ultimately providing real-time situational awareness in future threat monitoring systems. Future suggested research integrates this innovative approach with multiple authoritative data sources to further advance AI-driven modern threat analysis.
Type
Thesis
Description
Series/Report No
Department
Organization
Identifiers
NPS Report Number
Sponsors
Funding
Format
Citation
Distribution Statement
Distribution Statement A. Approved for public release: Distribution is unlimited.
Rights
This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.