Definition
Approximate Nearest Neighbors (ANN) is a class of optimization algorithms used in vector databases to retrieve high-dimensional vectors similar to a query point by sacrificing exhaustive search precision for significantly reduced latency and memory consumption.
Optimization for retrieval speed in vector databases, not to be confused with 'Artificial Neural Networks' or exact k-NN.
"Finding a specific book in a massive library by jumping to the right genre and shelf section instead of scanning every single spine on every floor."
- HNSW (Hierarchical Navigable Small World)(Component)
- Vector Embeddings(Prerequisite)
- Cosine Similarity(Metric)
- Product Quantization(Component)
Conceptual Overview
Approximate Nearest Neighbors (ANN) is a class of optimization algorithms used in vector databases to retrieve high-dimensional vectors similar to a query point by sacrificing exhaustive search precision for significantly reduced latency and memory consumption.
Disambiguation
Optimization for retrieval speed in vector databases, not to be confused with 'Artificial Neural Networks' or exact k-NN.
Visual Analog
Finding a specific book in a massive library by jumping to the right genre and shelf section instead of scanning every single spine on every floor.