Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Double Spike Dirichlet Priors for Structured Weighting
Huiming Lin, Meng Li
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 double spike Dirichlet prior instead of symmetric Dirichlet | Bayesian structured weighting · Simulation Scenario 1 | l1 error: 0.428 -> 0.032 | Improved |
| use joint double spike Dirichlet instead of two-step lasso | Structured simplex estimation · Simulation Scenario 1 | l1 error: 0.139 -> 0.032 | Improved |
| use double spike Dirichlet instead of two-step lasso | Structured simplex estimation · Simulation Scenario 2 | l1 error: 0.298 -> 0.210 | Improved |
| use double spike Dirichlet instead of symmetric Dirichlet | Structured simplex estimation · Simulation Scenario 2 | l1 error: 0.436 -> 0.210 | Improved |
| larger rho1 and smaller rho2 (ablation) | Double spike Dirichlet · Scenario 1 sensitivity | l1 error: l1 error decreases monotonically, indistinguishable for large rho1 | Improved |
| larger rho1 and smaller rho2 (ablation) | Double spike Dirichlet · Scenario 2 sensitivity | l1 error: monotone decreasing pattern no longer holds | No clear effect |
| fixed theta=1/K instead of Beta prior on theta (ablation) | Double spike Dirichlet · Simulation, deviation levels | l1 error: 0.420 vs 0.432 at deviation 0.3155 | Improved |
| two-step horseshoe instead of two-step lasso | Two-step baseline · Simulation, deviation levels | l1 error: 0.464 -> 0.453 at deviation 0.3155 | Improved |
| increase signal-to-noise ratio from 0.6 to 2.0 | Double spike Dirichlet · Simulation SNR | l1 error: 0.366 -> 0.017 | Improved |
| double spike Dirichlet instead of simple average | Forecast combination · ECB SPF GDP forecast | out-of-sample RMSE: 1.502 -> 1.441 | Improved |
| double spike Dirichlet instead of two-step lasso | Forecast combination · ECB SPF GDP forecast | out-of-sample RMSE: 1.542 -> 1.441 | Improved |
| random theta instead of fixed theta (ablation) | Double spike Dirichlet · ECB SPF GDP forecast | out-of-sample RMSE: 1.441 vs 1.444 | No clear effect |
Closest papers by meaning
- Scalable Bayesian Learning with posteriors (ICLR 2025 · similarity 0.86)
- Bayesian Neural Networks with Weight Sharing Using Dirichlet Processes (TPAMI 2018 · similarity 0.86)
- Sequential Monte Carlo Learning for Time Series Structure Discovery (ICML 2023 · similarity 0.86)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (NeurIPS 2024 · similarity 0.85)
- MEPSI: An MDL-Based Ensemble Pruning Approach with Structural Information (AAAI 2024 · similarity 0.84)
- HDT: Hierarchical Discrete Transformer for Multivariate Time Series Forecasting (AAAI 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.
