Chloe WilliamsIf you're looking to learn the fundamentals of vector databases, check out Zilliz's top 5 videos of 2023. Let's countdown, starting at number 5.
5: IVF Vector Index
IVF, or Inverted File index, is probably the most intuitive vector indexing method from a machine learning standpoint. It asks us, how can we easily cluster these points? and comes up with K-Means. When we query, we first look for the closest centroids, then into their clusters, massively reducing the number of computations from a brute for method.
4: ANNOY Vector Index
ANNOY stands for Approximate Nearest Neighbors, Oh Yeah. It's a vector search algorithm that came out of Spotify. It is a rougher version of IVF. It uses hyperplanes to divide the space in half between two points, and does this over and over again. When we query, we are essentially querying a binary tree.
3: How to Add Conversational Memory to an LLM Using LangChain
One of the main differences between talking to a person and talking to ChatGPT is memory. People have memories of their conversations. You can give this memory to ChatGPT using LangChain with a vector database. In this tutorial, we cover how.
2: Persistent Vector Storage with LlamaIndex
Learn how to use Milvus as persistent vector storage with LlamaIndex in under 5 minutes.
1: HNSW Vector Index
Hierarchical Navigable Small Worlds is a graph based index. When you are inserting the points, they are placed in some layers of the graph based on a uniform random number variable. At query time, the layers are queried top down.
If you want to keep up with the latest videos, subscribe to Zilliz's YouTube channel.
