Can I read Optimal Dimensionality Reduction for Sparse Vectors on EtoBox?
Optimal Dimensionality Reduction for Sparse Vectors by mittasnehanjali is a document available to read on EtoBox.
What is Optimal Dimensionality Reduction for Sparse Vectors about?
This paper investigates dimensionality reduction for s-sparse vectors, focusing on average-case guarantees and non-negative sparse vectors. It establishes that the birthday paradox map is optimal for average-case dimensionality reduction, requiring at least O(s^2) dimensions for preservation of norms in many scenarios. Additionally, it presents improved upper bounds for non-negative sparse vectors, allowing for smaller embeddings while preserving distances in various ℓp norms.
- Author
- mittasnehanjali
- Language
- EN