Embedding Atlas
Compute & interactively visualize large embeddings
108 followers
Compute & interactively visualize large embeddings
108 followers
Embedding Atlas is a tool that provides interactive visualizations for large embeddings. It allows you to visualize, cross-filter, and search embeddings and metadata. Open sourced by Apple.
Raycast
Looks like Apple is letting their research publish open source projects! If so — that's great news for the OSS ecosystem.
Embedding Atlas Features:
🏷️ Automatic data clustering & labeling: Interactively visualize and navigate overall data structure.
🫧 Kernel density estimation & density contours: Easily explore and distinguish between dense regions of data and outliers.
🧊 Order-independent transparency: Ensure clear, accurate rendering of overlapping points.
🔍 Real-time search & nearest neighbors: Find similar data to a given query or existing data point.
🚀 WebGPU implementation (with WebGL 2 fallback): Fast, smooth performance (up to few million points) with modern rendering stack.
📊 Multi-coordinated views for metadata exploration: Interactively link and filter data across metadata columns.
Check out the demo — impressive this is all done in the browser!
Promomix
@chrismessina Embedding Atlas addresses the friction of embedding visualizations head-on. According to its accompanying paper, it's designed for ease-of-use with minimal data wrangling, great scalability (millions of points!), and multi-view metadata exploration. It's rare to get that level of polish combined with real-time interactivity in a browser. Kudos to the team!
Super Intern
Didn't expect Apple to open-source something like this😂 Lowkey thought they'd keep it locked in the walled garden!
But def a welcome surprise. How’s the performance with ultra-large embeddings (like 1M+ dimensions)? And any plans for a native macOS app?
Super Intern
Very cool UI for data scientist. Since I used nomic atlas, I was fascinated by this efficient way to explore large-scale dataset. Fantastic work by Apple!
What To Build
Very cool interactive visualization for large embeddings is hugely valuable for debugging and exploration. Curious about performance on million‑vector scales and support for custom distance metrics.
We’re debuting today as well keen to hear what you think.
GPT-4o
Whoa, this Embedding Atlas is pretty awesom! The real-time search and nearest neighbors feature is truely a game changer – finding similar data points on the fly, directly in the visualization? That's super helpful for exploring large datasets, right? Kinda genius imo. How customizable is it for different embedding dimensions though?
WhiteSpace
I work with large embedding datasets—this visualization saves hours. Can it handle real-time updates or streaming data sources?
AutoAE
Almost exactly what I need now for my PhD!!!