Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Prior Specification for Bayesian Matrix Factorization via Prior Predictive Matching
Eliezer de Souza da Silva, Tomasz Kuśmierczyk, Marcelo Hartmann, Arto Klami
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 |
|---|---|---|---|
| Increase overdispersion (smaller NB r) in the data generating process | Analytic K estimate for PMF · Artificial data, negative binomial overdispersion | relative error of K | Hurt |
| Increase zero-inflation (lower p_obs) in the data | Analytic K estimate for PMF · Artificial data, sparsity | relative error of K: error stays below 25% after dropping 10% of observations | Hurt |
| Increase column scaling gamma_max (inconsistent margins) | Analytic K estimate for PMF · Artificial data, K=50 | estimate of K | Hurt |
| Use Bayesian optimization directly on PSIS-LOO instead of prior predictive matching | Poisson MF hyperparameter selection · hetrec-lastfm user-artists | PSIS-LOO: roughly an hour per run; almost half of runs worse than the instant solution | Mixed |
| Swap prior family (Beta, Half-Normal, Log-Normal instead of Gamma) (ablation) | Analytic K estimate for PMF · Artificial data | estimated K vs true K | No clear effect |
Closest papers by meaning
- Probabilistic Matrix Factorization for Automated Machine Learning (NeurIPS 2018 · similarity 0.88)
- Bayesian Extensive-Rank Matrix Factorization with Rotational Invariant Priors (NeurIPS 2023 · similarity 0.88)
- Empirical Priors for Prediction in Sparse High-dimensional Linear Regression (JMLR 2020 · similarity 0.87)
- A Particle-Based Variational Approach to Bayesian Non-negative Matrix Factorization (JMLR 2019 · similarity 0.87)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (NeurIPS 2024 · similarity 0.86)
- Scalable Bayesian Learning with posteriors (ICLR 2025 · similarity 0.85)
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.
