Need of Vector Databases for LLMs
Learn about types of vector embeddings and vector database with its working for LLMs.
We'll cover the following...
Types of embeddings
Embeddings can be applied to different types of raw data, including text, images, audio, and video.
Text embeddings: Imagine you have a box for words. Similar words, like "happy" and "joyful," get placed close together in the box because they're similar. We use natural language processing (NLP) word embeddings such as word2vec, GloVe, and FastText to represent words as a vector.
Image embeddings: This model is for pictures. For example, an image embedding for a cat might capture features like whiskers, pointed ears, and fur texture, while a dog embedding might capture floppy ears, a wagging tail, and a furry coat. These embeddings allow algorithms to find similar images based on these shared features. We use convolutional neural networks (CNNs) to extract image features, which are then converted into vector embeddings.
Audio embeddings: This model is for sounds. Similar sounds, like different instruments or animal noises, might be placed close together because they have similar qualities. ...