Approximate Vector Search with KMeans and Azure SQL | Data Exposed

In this episode, we'll see how to calculate KMeans clusters for vector data which will then be used to perform an Approximate Similarity Search. We'll offload resource intensive processing to calculate KMeans using SciKit-Learn to a container and then do cell probing in pure T-SQL.

Watch on Data Exposed

Resources:

Intelligent applications with Azure SQL Database: https://aka.ms/sqlai

Azure SQL Devs’ Corner: https://devblogs.microsoft.com/azure-sql/

Vector Search Optimization via KMeans, Voronoi Cells and Inverted File Index (aka “Cell-Probing”): https://devblogs.microsoft.com/azure-sql/vector-search-optimization-via-kmeans-voronoi-cells-and-inverted-file-index-aka-cell-probing/

View/share our latest episodes on Microsoft Learn and YouTube!