Research map / Learning theory and algorithms / Matrix and tensor methods
Multi-source Learning via Completion of Block-wise Overlapping Noisy Matrices
Doudou Zhou, Tianxi Cai, Junwei Lu
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 |
|---|---|---|---|
| use BONMI instead of pre-training on one source then continuing | embedding learning · Chinese-CUI relatedness, set I | rank correlation: 0.406 -> 0.671 | Improved |
| use BONMI instead of ALS | matrix completion · Chinese-CUI similarity, set I | rank correlation: 0.569 -> 0.611 | Improved |
| use BONMI instead of SMC | matrix completion · Chinese-CUI similarity, set II | rank correlation: 0.365 -> 0.707 | Improved |
| use BONMI instead of VGD | matrix completion · Chinese-CUI relatedness, set II | rank correlation: 0.385 -> 0.666 | Improved |
| use BONMI instead of VGD on within-corpus pairs | matrix completion · Chinese-Chinese relatedness, set I | rank correlation: 0.761 -> 0.741 | No clear effect |
| use BONMI instead of ALS on within-corpus pairs | matrix completion · CUI-CUI relatedness, set I | rank correlation: 0.689 -> 0.678 | No clear effect |
| use BONMI instead of ALS | translation · Chinese-to-CUI translation precision@5 | precision: 0.324 -> 0.398 | Improved |
| use BONMI instead of ALS | translation · Chinese-to-CUI translation precision@20 | precision: 0.485 -> 0.554 | Improved |
| increase observation rate p0 | all completion methods · Simulation setting (i) | relative error | Improved |
| increase number of sources m | BONMI · Simulation setting (ii) | relative error | Improved |
| increase number of sources m | pre-training and VGD · Simulation setting (ii) | relative error | No clear effect |
| increase noise level sigma from 0.3 to 0.5 | BONMI, ALS, Pre-train · Simulation setting (iii) translation | translation precision | Hurt |
Closest papers by meaning
- High Rank Matrix Completion With Side Information (AAAI 2018 · similarity 0.85)
- Scalable Completion of Nonnegative Matrix with Separable Structure (AAAI 2016 · similarity 0.84)
- Knowledge Graph Completion via Complex Tensor Factorization (JMLR 2017 · similarity 0.84)
- Optimal Transfer Learning for Missing Not-at-Random Matrix Completion (ICML 2025 · similarity 0.84)
- Matrix Completion with Incomplete Side Information via Orthogonal Complement Projection (ICML 2025 · similarity 0.83)
- A Pairwise Pseudo-likelihood Approach for Matrix Completion with Informative Missingness (NeurIPS 2024 · similarity 0.83)
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.
