Vector Database Concepts

Question 1

What is a vector database?

Answer 1

A specialized database designed to store and query high-dimensional vector embeddings.

Question 2

What is a vector embedding?

Answer 2

A numeric representation of data (text, images, etc.) in a high-dimensional space.

Question 3

Why use vector databases?

Answer 3

To perform efficient similarity search across large sets of embeddings.

Question 4

What is a similarity search?

Answer 4

Finding items that are most similar to a given query vector.

Question 5

What is cosine similarity?

Answer 5

A metric that measures the cosine of the angle between two vectors.

Question 6

What is Euclidean distance?

Answer 6

A metric that measures the straight-line distance between two points in space.

Question 7

What is dot product similarity?

Answer 7

A similarity metric based on the dot product of two vectors.

Question 8

What is an ANN index?

Answer 8

Approximate Nearest Neighbor index used to speed up similarity search.

Question 9

What is brute-force search?

Answer 9

A similarity search method that compares every vector — accurate but slow.

Question 10

What is HNSW?

Answer 10

Hierarchical Navigable Small World — a popular ANN algorithm.

Question 11

What is IVF in vector search?

Answer 11

Inverted File Index — used to partition vectors for faster search.

Question 12

What is PQ (Product Quantization)?

Answer 12

A technique to compress and speed up vector similarity searches.

Question 13

What is Faiss?

Answer 13

A library developed by Facebook for efficient similarity search and clustering of dense vectors.

Question 14

What is Milvus?

Answer 14

An open-source vector database for scalable similarity search.

Question 15

What is Weaviate?

Answer 15

A vector search engine with integrated machine learning capabilities.

Question 16

What is Pinecone?

Answer 16

A fully managed vector database service for production-grade applications.

Question 17

What is Qdrant?

Answer 17

An open-source vector search engine focused on performance and reliability.

Question 18

What is Vespa?

Answer 18

A platform for serving vector search and recommendation systems.

Question 19

What is vector dimensionality?

Answer 19

The number of features or coordinates in each embedding vector.

Question 20

What is vector normalization?

Answer 20

Scaling vectors to have unit length for consistent similarity metrics.

Question 21

What is index building?

Answer 21

The process of preparing vectors for fast similarity search.

Question 22

What is index probing?

Answer 22

Querying an ANN index to find nearest vectors.

Question 23

What is filtering in vector search?

Answer 23

Restricting search to vectors that meet certain metadata criteria.

Question 24

What is metadata in vector databases?

Answer 24

Structured attributes associated with each vector for filtering.

Question 25

What is hybrid search?

Answer 25

Combining keyword-based search with vector similarity search.

Question 26

What is semantic search?

Answer 26

Search based on meaning and context rather than exact keywords.

Question 27

What is multimodal search?

Answer 27

Search that supports multiple input types, like text and images.

Question 28

What is a collection in a vector database?

Answer 28

A logical grouping of vectors with similar purpose.

Question 29

What is a vector store?

Answer 29

A system or component that holds and manages vector embeddings.

Question 30

What is vector recall?

Answer 30

The proportion of relevant vectors successfully retrieved.

Question 31

What is vector precision?

Answer 31

The proportion of retrieved vectors that are relevant.

Question 32

What is embedding drift?

Answer 32

When the meaning of embeddings changes due to retraining.

Question 33

What is index rebalancing?

Answer 33

Redistributing data in an index for optimal performance.

Question 34

What is sharding in vector databases?

Answer 34

Splitting data across multiple machines for scalability.

Question 35

What is replication in vector databases?

Answer 35

Copying data across nodes for fault tolerance.

Question 36

What is vector ingestion?

Answer 36

The process of adding new embeddings to the vector database.

Question 37

What is vector deletion?

Answer 37

Removing vectors from a vector database.

Question 38

What is vector update?

Answer 38

Replacing or modifying vectors in the database.

Question 39

What is cold start in vector search?

Answer 39

Performance issues when index is not yet warmed up.

Question 40

What is ANN accuracy vs. speed tradeoff?

Answer 40

Faster searches may be less accurate due to approximation.

Question 41

What is latency in vector search?

Answer 41

The time it takes to return search results.

Question 42

What is throughput in vector search?

Answer 42

The number of queries a system can handle per second.

Question 43

What is batch querying?

Answer 43

Submitting multiple queries at once for efficiency.

Question 44

What is vector quantization?

Answer 44

Compressing vectors for faster processing and reduced memory usage.

Question 45

What is a vector similarity threshold?

Answer 45

A cutoff value to decide which vectors are similar enough.

Question 46

What is top-k retrieval?

Answer 46

Returning the top k most similar vectors for a given query.

Question 47

What is dynamic indexing?

Answer 47

Updating the index as new data is added or removed.

Question 48

What is recall@k?

Answer 48

The fraction of relevant items found in the top-k results.

Question 49

What is ANN benchmarking?

Answer 49

Evaluating accuracy and performance of approximate nearest neighbor algorithms.