Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
A Particle-Based Variational Approach to Bayesian Non-negative Matrix Factorization
Muhammad A Masood, Finale Doshi-Velez
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 |
|---|---|---|---|
| initialize NMF with Q-Transform instead of random restarts | particle-based Bayesian NMF · NMF benchmark datasets (M=5) | runtime to produce factorizations | Improved |
| use Q-Transform initializations instead of NNDSVDar | particle-based Bayesian NMF · NMF benchmark datasets | Stein discrepancy (posterior quality) | Improved |
| use Q-Transform particles instead of thinned HMC samples | Bayesian NMF posterior approximation · Olivetti Faces, LFW, Faces BIO (M=5,25,50) | Stein discrepancy | Improved |
| use SVGD or DSGD gradient optimization instead of initialization-based particles | Bayesian NMF posterior approximation · NMF benchmark datasets (M=5) | Stein discrepancy and factorization quality | Hurt |
| increase observation noise epsilon in synthetic data (ablation) | Q-Transform initialization · synthetic data D=N=500, K=20 | reconstruction error ratio and time ratio vs random restart | Hurt |
| increase transfer rank R_T from small values up to 10 (ablation) | Q-Transform initialization · synthetic transfer to 500x500 rank-10 data | initialization quality | Hurt |
| apply Q-Transform with partially observed data (10%, 30%, 50% missing) | Bayesian NMF · Olivetti Faces with missing data | Stein discrepancy | Improved |
Closest papers by meaning
- Analysis of Variational Bayesian Factorizations for Sparse and Low-Rank Estimation (ICML 2016 · similarity 0.89)
- Bayesian group factor analysis with structured sparsity (JMLR 2016 · similarity 0.88)
- Diversified Bayesian Nonnegative Matrix Factorization (AAAI 2020 · similarity 0.88)
- Bayesian Extensive-Rank Matrix Factorization with Rotational Invariant Priors (NeurIPS 2023 · similarity 0.88)
- Multilayer Matrix Factorization via Dimension-Reducing Diffusion Variational Inference (ICML 2025 · similarity 0.87)
- Covariate-moderated Empirical Bayes Matrix Factorization (NeurIPS 2025 · similarity 0.86)
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.
