Xilinx Recommendation Engine for TigerGraph 3.x Overview

An application built on TigerGraph consists of application software, TigerGraph GSQL queries, and possibly also user-defined functions (UDFs).

The Xilinx Recommendation Engine for TigerGraph 3.x adds to the application with a plug-in module of UDFs that interface with the Xilinx Maximal Independent Set Alveo Product. The Cosine Similarity Alveo Product, in turn, interfaces with the Alveo accelerator card to provide hardware acceleration of cosine similarity operations, as shown in the diagram below.

By writing queries that call the Recommendation Engine UDFs, your application can accelerate cosine similarity operations using a Xilinx Alveo accelerator card. Please see Xilinx Cosine Similarity Alveo Product Overview for an overview of what cosine similarity is and how you can use cosine similarity on an Alveo accelerator card to find patterns in your data quickly.

The following sections walk you through the installation of the Recommendation Engine, the operation and composition of a demonstration application, and how to write your own TigerGraph-based application that uses the Recommendation Engine.