Vector Databases and Semantic Search Implementation
Chapter 1: Embeddings and Vector Spaces
From Data to Vectors: A Refresher
Survey of Embedding Models
Understanding Vector Dimensionality
Introduction to Dimensionality Reduction
Measuring Similarity in Vector Space
Hands-on Practical: Generating and Comparing Embeddings
Quiz for Chapter 1
Chapter 2: Introducing Vector Databases
What Defines a Vector Database?
Core Architectural Components
Data Models and Schemas
Vector Operations: CRUD
Metadata Filtering
Scaling Considerations
Hands-on Practical: Basic Vector DB Interaction
Quiz for Chapter 2
Chapter 3: Approximate Nearest Neighbor (ANN) Search
The Need for Approximation
Core Concepts of ANN
Algorithm Overview: HNSW
Algorithm Overview: IVF
Algorithm Overview: LSH
Indexing Parameters and Tuning
Evaluating ANN Performance
Hands-on Practical: Experimenting with Index Parameters
Quiz for Chapter 3
Chapter 4: Building Semantic Search Systems
Semantic vs. Keyword Search Revisited
Architecture of a Semantic Search Pipeline
Data Preparation and Chunking Strategies
Query Processing and Embedding
Result Ranking and Re-ranking
Implementing Hybrid Search
Evaluating Semantic Search Relevance
Hands-on Practical: Designing a Search Query Flow
Quiz for Chapter 4
Chapter 5: Vector Databases in Practice
Choosing a Vector Database Platform
Working with Pinecone Client
Working with Weaviate Client
Working with Milvus Client
Working with ChromaDB Client
Indexing Large Datasets Efficiently
Monitoring and Maintenance
Hands-on Practical: Build a Small Semantic Search App
Quiz for Chapter 5
Metadata Filtering
Was this section helpful?
- Filtering in Qdrant: How it works and why it’s important, Kacper Łukawski, 2022 (Qdrant) - Detailed explanation of metadata filtering strategies (pre-filtering, post-filtering) and their implications in a production vector database.
- Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs, Yu. A. Malkov, D. A. Yashunin, 2018 IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 40 (IEEE) DOI: 10.1109/TPAMI.2017.2762879 - Describes the core HNSW algorithm used by many vector databases, providing context for how filtering can be integrated or applied over such structures. Originally published as a pre-print on arXiv.
- Filtering Nearest Neighbors for Hybrid Search, Haoxiang Li, Yandi Li, Minjia Zhang, Song Liu, Jian Li, 2023 Proceedings of the VLDB Endowment, Vol. 16 (VLDB Endowment) DOI: 10.14778/3611477.3611502 - Discusses advanced techniques for efficiently integrating metadata filtering into approximate nearest neighbor search, covering various strategies and their performance implications.