Definition
The fixed number of scalar values in a vector that represent the semantic features of a data point. In RAG architectures, it defines the size of the latent space where text chunks are mapped, directly impacting the granularity of semantic search and the memory requirements of the vector store.
Not the total count of vectors in a database, but the numerical length of a single vector.
"The number of 'GPS coordinates' required to find a specific thought in a high-dimensional library; more coordinates provide a more precise location but require a larger map."
- Vector Database(Storage Infrastructure)
- Cosine Similarity(Mathematical Metric)
- Matryoshka Embeddings(Optimization Technique)
- Quantization(Compression Method)
Conceptual Overview
The fixed number of scalar values in a vector that represent the semantic features of a data point. In RAG architectures, it defines the size of the latent space where text chunks are mapped, directly impacting the granularity of semantic search and the memory requirements of the vector store.
Disambiguation
Not the total count of vectors in a database, but the numerical length of a single vector.
Visual Analog
The number of 'GPS coordinates' required to find a specific thought in a high-dimensional library; more coordinates provide a more precise location but require a larger map.