Intro to Vector Databases
Intro to Vector Databases
Databases designed to store and search data using vector embeddings, enabling efficient similarity search for unstructured data like text and images.Intro to Vector Databases
Vector
Vectors, or vector embeddings in databases, are quantities with magnitude and direction...Intro to Vector Databases
Vector Embedding
Numerical representations of objects, such as words or images, in a vector space...Search
A search method that combines vector search with traditional keyword search to improve retrieval accuracy and relevance.Search
Sparse Vectors
Sparse embeddings are generated from algorithms like BM25 and SPLADE...Search
Dense Vectors
In contrast to sparse vectors, dense vectors contain mostly non-zero values and are generated from machine learning models like GloVe and Transformers...Search
BM25/BM25F
BM25 is a ranking function used by search engines to estimate the relevance of documents to a given search query. It is part of the family of probabilistic information retrieval models...Hierarchical Navigable Small World
An indexing algorithm used in vector databases to enable fast and efficient similarity search.Hierarchical Navigable Small World
Hierarchical Graph Structure
HNSW utilizes a multi-layered graph structure where each layer is a graph of the data points. The top layers contain fewer points with long-range connections...Hierarchical Navigable Small World
ANN - Approximate Nearest Neighbor
ANN algorithms calculate the approximate nearest neighbors to a query, as opposed to kNN algorithms, which calculate the true nearest neighbors...Hierarchical Navigable Small World
Layered Navigation
During a search, HNSW starts at the topmost layer and makes 'big jumps' across the graph to quickly move closer to the target region...Multimodal RAG
A technique that combines retrieval of relevant multimodal data, such as images, text, audio, or video, with generative large language models to generate natural language responses or content to a query.Multimodal RAG
Cross Modal Reasoning
Cross-modal reasoning refers to the ability to make connections by integrating information from different modalities or sources including text, images, audio and video...Multimodal RAG
Multimodal Embeddings Models
Multimodal Embeddings Models produce a joint embedding space for multimodal data that understands text, images, audio and more...Multimodal RAG
Multimodal Contrastive Finetuning
Multimodal Embeddings Models produce a joint embedding space for multimodal data that understands text, images, audio and more...Databases
Systems for storing and storing, organizing, and retrieving structured or unstructured data efficiently.Databases
Graph Database
A graph database stores data in nodes and edges, representing entities and their relationships. It's optimized for...Databases
Inverted Indexes
An inverted index is a database indexing structure that maps keywords to their locations in documents...Databases
Sharding
Sharding is splitting a database into smaller, faster, more easily managed parts called shards...Large Language Models
Deep learning models trained on massive datasets to understand and generate human-like text, used in applications like chatbots and content generation.Large Language Models
Large Language Model (LLM)
A Large Language Model (LLM) is a machine learning model that is trained on vast text data, learning language...Large Language Models
Finetuning
Fine-tuning is a process where a pre-trained machine learning model is further trained on a specific dataset...Large Language Models
Multi-modal
Multi-modal learning involves integrating various data types like images, text, audio, and sensory inputs...Information Retrieval/Search
Techniques for finding relevant information in a large collection of data, such as documents, images, or videos.Information Retrieval/Search
Reranking
Re-ranking is adjusting the scores of search results after initial retrieval. It usually uses a machine learning model...Information Retrieval/Search
Retrieval Augmented Generation (RAG)
Retrieval Augmented Generation (RAG) is the process of contextualising prompts for large language models...Embedding Types
Different types of embeddings used in vector databases, such as text embeddings, image embeddings, and multimodal embeddings.Embedding Types
Variable Dimensions
Flexible embedding sizes, like Matryoshka embeddings. Encode information hierarchically, allowing adaptation...Embedding Types
Sparse Embeddings
Sparse vectors are often high-dimensional with many zero values. They are generated from algorithms like BM25 and SPLADE...Embedding Types
Quantized Embeddings
Compressed dense vectors using lower-precision data types (e.g., float32 to int8). Reduces memory usage and speeds up search...Chunking Techniques
Techniques for breaking down large data into smaller, more manageable chunks for processing and storage.Chunking Techniques
Semantic Chunking
In this technique, the text is divided into meaningful units, such as sentences or paragraphs, which are then vectorized...Chunking Techniques
Recursive Chunking
Text is initially split using a primary separator, like paragraphs. If the resulting chunks are too large, secondary separators...Chunking Techniques
LLM-Based Chunking
This advanced technique uses a Language Model (LLM) to generate chunks. The LLM processes the text and generates semantically isolated sentences...Advanced RAG Techniques
Techniques for improving the performance and capabilities of the RAG model, such as training on custom datasets, fine-tuning, and optimizing for specific tasks.Advanced RAG Techniques
Reasoning and Acting (ReAct)
ReAct prompting combines CoT with agents, creating a system where the model can generate thoughts and delegate actions to agents that interact with external data sources...Advanced RAG Techniques
Tree of Thoughts (ToT)
Tree of Thoughts prompting builds on CoT by instructing the model to evaluate its responses at each step in the problem-solving process or even generate several different solutions to a problem...Advanced RAG Techniques
