Apply Code UMAP 2022 - Dimensional Reduction of Embeddings w/ Python, Colab Jupyter NB

Our SBERT (BI-encoder) data live in a high-dim vector space. But we want a 3D visualization. Therefore we use Python Library UMAP - Uniform Manifold Approximation and Projection for Dimension Reduction!
An Introduction to dimensional reduction with classical UMAP, next video will focus on more advanced versions! Yes, finally! Topological manifolds.
All credits to:
umap-learn.readthedocs.io/en/...
arxiv.org/abs/1802.03426
by Leland McInnes, John Healy, James Melville
#topologicalspace
#datascience
#machinelearningwithpython
#embedding
#dimensional
#dimensions

[@Azariven]

Wow, so much more clearer with how UMAP can be implemented and used. I assume in the current world of LLM we can replace the tfidf vectorizer with one of those chatgpt vector embedding like text-embedding-ada-002 and get a much better separation?

[@surajitchakraborty1903]

Hi Thanks for the fantastic video. Are you able to share the Colab Notebook ?

[@code4AI]

Recommend you check this link to the author of UMAP umap-learn.readthedocs.io/en/latest/ or their GitHub repos for their code implementation, simply because of general SW copright information & their detailed license specifications.