From the course: Introduction to AI-Native Vector Databases
Join today to access over 24,400 courses taught by industry experts.
Ways to measure performance of a vector DB
From the course: Introduction to AI-Native Vector Databases
Ways to measure performance of a vector DB
In the last video, we spoke about the HNSW ANN algorithm, which is the core technology that allows vector databases to perform real-time searches over lots and lots of data. In this video, we'll introduce some concepts that will help you assess the performance of a vector database. Depending on your use case, some of these metrics might be more important than others. Going back to using ANN, this means that not all nearest neighbors might be returned. In order to measure how reliably and accurately nearest neighbors are returned, we use a metric called recall. Recall is defined as the number of retrieved nearest neighbors by ANN over the number of total nearest neighbors if there were no approximations. The higher the recall, the better. It means that our database is accurately retrieving similar objects to the query, and not missing any of the neighbors that should be returned. If we want to perform real-time search with our vector database, we'll have to compromise on recall to get…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
