Knowledge-Graphs on Ryan Orban

A Survey on Graph Representation Learning Methods

me@ryanorban.com (Ryan Orban) — Thu, 07 Jul 2022 00:00:00 +0000

Summary

Khoshraftar and An (York University) survey the field of graph representation learning, organizing methods into three broad families: shallow node embeddings, graph neural networks, and knowledge graph embeddings. The goal throughout is to learn low-dimensional vector representations that preserve graph structure — proximity, community membership, relational patterns — so that downstream tasks like node classification, link prediction, and graph classification can be solved with standard ML pipelines.

Graph-Powered Machine Learning

me@ryanorban.com (Ryan Orban) — Tue, 05 Apr 2022 00:00:00 +0000

Summary

Alessandro Negro’s 2021 Manning Publications book is a practitioner-oriented treatment of applying machine learning to and with graph-structured data. The book uses Neo4j as the primary graph database throughout, grounding abstract concepts in a system readers can run locally. Topics span knowledge graphs, graph neural networks (GNNs), community detection, and link prediction, forming a progression from graph fundamentals to advanced ML techniques.

An Introduction to Knowledge Graphs

me@ryanorban.com (Ryan Orban) — Sat, 22 May 2021 00:00:00 +0000

Summary

This Stanford AI Lab blog post introduces knowledge graphs — structured representations of real-world entities and their relationships — covering both the conceptual foundations and the practical challenges of building and maintaining them. A knowledge graph stores facts as triples: (entity, relation, entity), e.g., (Barack Obama, bornIn, Hawaii). The structure enables reasoning, question answering, and semantic search that plain text retrieval can’t match.