RAG: Graph Retrieval vs Graph Reasoning
Building Knowledge Graphs in RAG is not just about using graphs as data stores, but also as a means for structured reasoning.
Transformers themselves as a type of Graph Neural Network
It is widely recognised that Large Language Models, such as GPT, are built on the Transformer architecture. However, what may be less widely acknowledged is… | 83 comments on LinkedIn
A new class of GNNs grand + blend
A new class of GNNs! This Tuesday in the #graph reading group, James Rowbottom and Ben Chamberlain present their "GRAND: Graph Neural Diffusion" paper ...
Joins and NULLs in SPARQL - Stardog
Joins play a pretty important role for defining the semantics of evaluating SPARQL queries even though they are not a part of the SPARQL syntax. One does not have to think about joins as long as their queries are restricted to basic graph patterns. However once more complex constructs appear in the query, their results are typically combined using the good old relational join operator. It has certain quirks, for example, in how it deals with nulls, and it's important to understand those to avoid result explosion and performance problems.
Nike: A Social Graph at Scale with Amazon Neptune | Amazon Web Services
Getting a graph database to be performant and easy to use is very different from making a NoSQL (non-relational) database high-performing. Listen in as Todd Escalona of AWS talks with Marc Wangenheim, Senior Engineering Manager at Nike, about how the company powers a number of applications via a social graph, built on Amazon Neptune, which […]