Researchers have developed a new algorithm for Sparse Principal Component Analysis (SPCA) that addresses the complexity arising from overlapping supports of components. While the problem is NP-hard, even for simple cases with two components and fixed rank, the new approach offers a fully polynomial-time approximation scheme (FPTAS) when the number of components and the rank are fixed. This algorithm guarantees orthonormal columns in the solution and provides real algebraic numbers as entries. The method involves reducing SPCA to a sparse version of the orthogonal Procrustes problem, which is then solved using approximate dynamic programming and iterative updates. AI
RANK_REASON The item is a research paper detailing a new algorithm for a specific statistical problem. [lever_c_demoted from research: ic=1 ai=0.4]
Read on Hugging Face Daily Papers →
- covariance matrix
- dynamic programming
- Fully polynomial-time approximation scheme
- NP-hardness of Euclidean sum-of-squares clustering
- Orthogonal Procrustes problem
- Sparse PCA
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →