Research map / Learning theory and algorithms / Prediction and decision losses
GaSPing for Utility
Mengyang Gu, Debarun Bhattacharjya, Dharmashankar Subramanian
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 GaSP instead of linear interpolation | utility estimator · Synthetic noise-free n=7, alpha=0.7 | out-of-sample MSE: 9.5 × 10−6 -> 2.8 × 10−7 | Improved |
| use GaSP instead of linear interpolation | utility estimator · Synthetic noise-free n=7, alpha=2.5 | out-of-sample MSE: 3.8 × 10−4 -> 5.0 × 10−6 | Improved |
| use GaSP instead of exponential parametric fit | utility estimator · Synthetic noise-free n=7, alpha=1.5 | out-of-sample MSE: 3.2 × 10−4 -> 5.5 × 10−6 | Improved |
| use GaSP instead of QPD | utility estimator · Synthetic noise-free n=7, alpha=2.5 | out-of-sample MSE: 2.8 × 10−3 -> 5.0 × 10−6 | Improved |
| use GaSP with a misspecified mean basis (x^0.5) (ablation) | GaSP · Synthetic noise-free power utility | out-of-sample MSE | No clear effect |
| use GaSP instead of linear interpolation under additive noise | utility estimator · Synthetic noisy n=10, alpha=0.7 | AvgMSE: 1.8 × 10−5 -> 9.7 × 10−6 | Improved |
| use GaSP instead of linear interpolation under additive noise | utility estimator · Synthetic noisy n=10, alpha=2.5 | AvgMSE: 1.9 × 10−4 -> 1.3 × 10−4 | Improved |
| add noise to assessments (noise-free to noisy) | all methods · Synthetic power utility | MSE | Hurt |
| use GaSP second-derivative posterior for curvature instead of LI slope changes | curvature classification · Synthetic noisy exponential utilities, n=15 | concavity classification | Improved |
| use GaSP instead of linear interpolation | utility estimator · Single-attribute real data, loss domain | AvgMSE: 9.7 × 10−4 -> 8.9 × 10−4 | Improved |
| use GaSP instead of linear interpolation | utility estimator · Single-attribute real data, gain domain | AvgMSE: 8.2 × 10−5 -> 7.4 × 10−5 | Improved |
| use GaSP instead of power/exponential parametric fit | utility estimator · Single-attribute real data, loss domain | AvgMSE: 1.7 × 10−3 -> 8.9 × 10−4 | Improved |
Closest papers by meaning
- Learning GAI-Decomposable Utility Models for Multiattribute Decision Making (AAAI 2024 · similarity 0.83)
- DT-PBO: an Interpretable Tree-based Surrogate Model for Preferential Bayesian Optimization (NeurIPS 2026 · similarity 0.83)
- Bayesian Optimization with Preference Exploration using a Monotonic Neural Network Ensemble (NeurIPS 2025 · similarity 0.82)
- Model Alignment as Prospect Theoretic Optimization (ICML 2024 · similarity 0.82)
- Preference Learning of Latent Decision Utilities with a Human-like Model of Preferential Choice (NeurIPS 2024 · similarity 0.81)
- Personalized Decision Modeling: Utility Optimization or Textualized-Symbolic Reasoning (NeurIPS 2025 · similarity 0.81)
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.
