Research map / Probabilistic and scientific modeling / Variational and generative sampling
H-Likelihood Approach to Deep Neural Networks with Temporal-Spatial Random Effects for High-Cardinality Categorical Features
Hangbin Lee, Youngjo Lee
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 |
|---|---|---|---|
| exact h-likelihood MLE (negative h-likelihood loss) instead of block-diagonal approximation of integrated likelihood | LMMNN MLP · Simulated spatial data | test MSPE: HL beats LMMNN-R at all l2 values (e.g., 1.11 vs 1.29 at l2=0.1) | Improved |
| exact MLE of dispersion parameters via h-likelihood instead of block-diagonal approximation (ablation) | RBF length-scale estimation · Simulated spatial data | estimated l2: HL 0.0967 vs LMMNN-R 0.48 for true 0.1; 8.929 vs 4.93 for true 10 | Improved |
| REML instead of ML for dispersion parameters (ablation) | HL dispersion estimation · Simulated spatial data | estimated variance components: REMLE nearly identical to MLE (e.g., l2 0.0971 vs 0.0967) | No clear effect |
| adjustment of random effect predictors (constraint centering) (ablation) | random effect predictions · Simulated spatial data | RMSE of random effect predictors: 0.14 (0.06) without vs 0.13 (0.05) with adjustment | Improved |
| exact h-likelihood approach instead of LMMNN block-diagonal approximation | Air quality data · Asthma air quality | test MSPE: 0.023 vs 0.035 for LMMNN-R | Improved |
| exact h-likelihood approach instead of LMMNN block-diagonal approximation | Cars data · Asthma cars | test MSPE: 0.084 for both | No clear effect |
| exact h-likelihood approach instead of single-random-effect LMMNN | Income data · Asthma income | test MSPE: 0.028 vs 0.027 for LMMNN-E (comparable) | Hurt |
Closest papers by meaning
- Integrating Random Effects in Deep Neural Networks (JMLR 2023 · similarity 0.88)
- Leveraging the Exact Likelihood of Deep Latent Variable Models (NeurIPS 2018 · similarity 0.86)
- How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model (ICML 2024 · similarity 0.86)
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · similarity 0.85)
- Likelihood Training of Cascaded Diffusion Models via Hierarchical Volume-preserving Maps (ICLR 2024 · similarity 0.85)
- A Likelihood Based Approach to Distribution Regression Using Conditional Deep Generative Models (ICML 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.
