Research map / Learning theory and algorithms / Matrix and tensor methods
A Non-commutative Extension of Lee-Seung's Algorithm for Positive Semidefinite Factorizations
Yong Sheng Soh, Antonios Varvitsiotis
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Run MMU with 50 random initializations for 500 iterations on a 20x20 random distance matrix with rank-2 PSD factorization | MMU algorithm · Random Euclidean distance matrix (n=20) | normalized squared error | Improved |
| Use a 7-dimensional full PSD factorization instead of a 27-dimensional NMF to decompose a face image | MMU algorithm · CBCL Face Database | interpretability of learned building blocks | Hurt |
| Use a block-diagonal PSD factorization with 9 blocks of size 2 instead of a full 7-dimensional PSD factorization (ablation) | MMU algorithm · CBCL Face Database | locality of learned building blocks | Improved |
| Learn a continuum of atoms from a single 2x2 block instead of finitely many NMF atoms | MMU block-diagonal PSD factorization · CBCL Face Database | expressiveness of geometry captured | Improved |
| Initialize MMU with diagonal PSD matrices | MMU algorithm · Theory (Section on NMF link) | iterate structure | Improved |
Closest papers by meaning
- Fixed-Rank Approximation of a Positive-Semidefinite Matrix from Streaming Data (NeurIPS 2017 · similarity 0.87)
- Structured Semidefinite Programming for Recovering Structured Preconditioners (NeurIPS 2023 · similarity 0.86)
- Constrained Factorization with Diagonal Scaling: Rank-Revealing Training and Pruning (NeurIPS 2026 · similarity 0.86)
- Dropping Symmetry for Fast Symmetric Nonnegative Matrix Factorization (NeurIPS 2018 · similarity 0.86)
- Sublinear Time Quantum Sensitivity Sampling (NeurIPS 2026 · similarity 0.84)
- The Structural Complexity of Matrix-Vector Multiplication (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
