08-embeddings

Embeddings

Word Vectors and the Universal Sentence Encoder

• 📝 Understanding word vectors
• 🚂 What is word2vec, 🚂 Color Vectors
• 📚 2013 paper Efficient Estimation of Word Representations in Vector Space
• 📚 2018 Universal Sentence Encoder paper

Embeddings

• 🎨 Embeddings Projector and Visualizing High Dimensional Space, Atlas
• 📝 Embeddings tutorial from Cohere
• 📚 What are embeddings? by Vicki Boykis
• 📚 Embeddings: What they are and why they matter
• 📚 Using open-source models for faster and cheaper text embeddings

Similarity

• 📚 Similarity between words and sentences
• 🎥 Cosine Similarity from StatQuest

Retrieval Augmented Generation (RAG)

• 📚 How to use retrieval augmented generation

Code Examples with Replicate

• 💻 Making your own Embeddings "Database", uses bge-large-en-v1.5
• 💻 Retrieval Augmented Generation (RAG) with p5.js + Replicate

On device, in browser embeddings (no node.js required!!)

• 💻 Embeddings with Transformers.js, uses Xenova/all-MiniLM-L6-v2

Clustering with UMAP dimensionality reduction

• Understanding UMAP
• umap-js
• UMAP p5.js examples
  • Simple clustering 3D colors -> 2D
  • Simple clustering random data, adjusting umap parameters
  • Animating UMAP process with larger color dataset
  • Clustering sentence embeddings from a database!

Assignment

• 🎨 Prepare to present a final project proposal.