P
PAIR Code
pair-code.github.io · Company blog · on HN since 2017
PAIR Code is Google's People + AI Research group, publishing interactive explainers and tools for understanding machine learning models and data.
What HN saysHN links it for tools like Facets and the interactive 'Understanding UMAP' article, which commenters use for exploring datasets and for a UMAP versus t-SNE discussion.
27HN people shared it
9linked it in comments
25stories submitted
281story points
6 months agolast shared
2017people per year2026
Most-shared posts
- 1Understanding UMAP (2019)8 people · 45 points · first shared 2019 · 17 comments on HN
- 2Facets / Visualizations for machine learning datasets6 people · 14 points · first shared 2017 · discussion on HN
- 3A hardware-accelerated machine intelligence library for the web4 people · 79 points · first shared 2017 · 7 comments on HN
- 4What-If Tool3 people · 4 points · first shared 2018 · discussion on HN
- 5What-If Tool: inspect machine learning models2 people · 4 points · first shared 2018 · discussion on HN
- 6Patchscopes: A Unifying Framework for Inspecting Hidden Representations of LMs2 people · 3 points · first shared 2024 · discussion on HN
- 7Language, trees, and geometry in neural networks1 people · 125 points · first shared 2019 · 5 comments on HN
- 8Explore Covid-19 Symptoms Search Trends1 people · 2 points · first shared 2020 · 1 comments on HN
- 9Google: Deep learning for JS1 people · 2 points · first shared 2017 · discussion on HN
- 10Language Interpretability Tool for visualization and understanding of NLP models1 people · 2 points · first shared 2021 · discussion on HN
- 11Communicating Model Uncertainty over Space1 people · 1 points · first shared 2020 · discussion on HN
- 12Font explorer1 people · first shared 2023
- 13Quickdraw1 people · first shared 2017
What HN people wrote when they linked it
cannoneyed · 2021-07-03
Shameless self plug here, but I saw this today and figured I should link to an interactive overview article "Understanding UMAP" my colleagues and I put together:
https://pair-code.github.io/understanding-umap/
UMAP is a really useful piece in the modern data science toolkit, and despite its power it's surprisingly simple and elegant. But as with all dimensionality reduction techniques, there's a lot of ways to misread the results. High dimensional data behaves very counterintuitively, and any reduction in dimensionality fundamentally distorts the original data in some way. Understanding the fundamental concepts behind UMAP and exploring how it works is the best way to develop an intuition of what the technique can and can't tell you about your data.
mncharity · 2017-08-11
> where would you recommend starting [emphasis mine]
Play with http://playground.tensorflow.org/ . Read today's https://news.ycombinator.com/item?id=14992865 about https://pair-code.github.io/deeplearnjs/ .
stared · 2025-10-15
I am curious - why UMAP not t-SNE? (See https://pair-code.github.io/understanding-umap/) When I saw the vis, there is a a collection of lines, which look as an artifact. t-SNE (typically) gives more "organic" results of blobs, provided you set perplexity high enough.
Also, while I admire examples of instances, it would be interesting to the map or original laguages - which is close to which, in terms of their English accents.
oops · 2023-10-05
I have been poring over a large amount of data for debugging purposes recently and discovered “facets” which provides a quick and easy way to compare and explore datasets.
https://pair-code.github.io/facets/
mortenjorck · 2023-08-17
The underlying UMAP model is actually pretty interesting. It's linked to in the tour, though I would have expected it to be featured more prominently: https://pair-code.github.io/understanding-umap/
leobg · 2023-03-11
I've got 2 questions:
1) Can I run this, say, on Colab? I see no code.
2) Any interactive open source tool to visually explore the latent space of a model’s representations for any given prompt, or for faces, buildings, etc.? Like they exist for fonts [1].
[1] https://pair-code.github.io/font-explorer/
MrPoufPouf · 2020-02-08
Sounds like what What-It Tool (https://pair-code.github.io/what-if-tool/) provides.
nl · 2017-10-09
The front page https://pair-code.github.io/facets/ is more useful, because you can load up your own data and try it out.
elsherbini · 2017-10-08
They have an example of using the tool on the quick draw dataset: https://pair-code.github.io/facets/quickdraw.html