Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Inference in Mixed Probabilistic Submodular Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 mixed FLDC (both repulsive and attractive higher-order potentials) instead of FLID alone | product recommendation model · Amazon baby registry, 13 categories | AUC (posterior marginals): average 0.766327 vs 0.725653 for FLID | Helped |
| Add attractive (log-supermodular) potentials to diversity model, FLDC vs FLID (ablation) | FLID · Amazon baby registry, datasets with at most 62 items | AUC: e.g., furniture 0.759979 vs 0.739646 | Helped |
| Use mixed FLDC instead of modular model | product recommendation model · Amazon baby registry, 13 categories | AUC (posterior marginals): average 0.766327 vs 0.708698 | Helped |
| Add repulsive (log-submodular) higher-order potentials for small ground sets (ablation) | FLID vs modular · safety, furniture, carseats | accuracy | Helped |
| Use FLIC (attractive only) instead of FLID (repulsive only) on larger ground sets | product recommendation model · categories with more than 40 items (except health) | accuracy | Helped |
| Use FLIC instead of FLID on small ground sets | FLIC vs FLID · safety, furniture, carseats | accuracy | Hurt |
| Mixed FLDC vs FLIC on diaper (AUC) | FLDC vs FLIC · diaper, Amazon baby registry | AUC: 0.787274 vs 0.787457 | No clear effect |
| Vary latent dimension split L and K at fixed L+K=10 (mixed both nonzero vs single type) (ablation) | FLDC · diaper, accuracy | accuracy | Helped |
Closest papers by meaning
- Provable Variational Inference for Constrained Log-Submodular Models (NeurIPS 2018 · similarity 0.93)
- Optimal Continuous DR-Submodular Maximization and Applications to Provable Mean Field Inference (ICML 2019 · similarity 0.90)
- Robust, Accurate Stochastic Optimization for Variational Inference (NeurIPS 2020 · similarity 0.89)
- Universal Boosting Variational Inference (NeurIPS 2019 · similarity 0.89)
- Surrogate Likelihoods for Variational Annealed Importance Sampling (ICML 2022 · similarity 0.89)
- Theoretical Guarantees for Variational Inference with Fixed-Variance Mixture of Gaussians (ICML 2024 · similarity 0.87)
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.