Researchers have analyzed the computational complexity of entrywise power matrix factorization (EPMF), a problem that includes modulus models and component-wise square factorization as special cases. They established a complete complexity landscape for both exact and approximate EPMF. For the exact case, they proved it is equivalent to the signing problem, which is strongly NP-hard but solvable in polynomial time for a fixed rank. In the approximate case, EPMF was shown to be NP-hard even for a rank of two. AI
IMPACT This research provides theoretical insights into matrix factorization techniques, which could inform future algorithm development in machine learning.
RANK_REASON The cluster contains a research paper detailing computational complexity analysis of a mathematical problem. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →