Research map / Probabilistic and scientific modeling / Variational and generative sampling
Importance Sampling Tree for Large-scale Empirical Expectation
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 |
|---|---|---|---|
| sample with IST using gradient norm instead of uniform | ANN · Synthetic 2D oscillating boundary | test error: 2.64 -> 0.62 | Helped |
| use loss per sample instead of gradient norm as importance (ablation) | ANN-IST · Synthetic 2D oscillating boundary | test error: 0.62 -> 1.58 | Hurt |
| IST-L vs uniform baseline | ANN-IST-L · Synthetic 2D oscillating boundary | test error: 2.64 -> 1.58 | Helped |
| resample uniformly for each stump (Boost-U) | Boost · Adaboost synthetic 2D | training/validation loss | Hurt |
| sample with IST for each stump (Boost-IST) | Boost · Adaboost synthetic 2D | validation loss | Helped |
| fixed uniform training set (Boost baseline) | Boost · Adaboost synthetic 2D | validation loss | Hurt |
| IST sampling instead of uniform | CNN · CIFAR10 | 0.75-quantile gradient norm: 2.64 × 10−4 -> 0.52 | Helped |
| number of epochs of training | IST CNN runs · CIFAR10 | weight correlation between runs: -0.017 -> 0.9897 | Helped |
| increase number of samplings T | IST SVM · Covertype | test error | Helped |
| IST instead of uniform sampling for SVM prediction | IST SVM · Covertype | test error vs inner products | Helped |
| apply IST to webspam or a9a | IST SVM · webspam, a9a | computation cost | Hurt |
| close-bound bifurcation policy vs mean and max | IST SVM · synthetic checkerboard | sampling aggressiveness | No clear effect |
Closest papers by meaning
- Variable Importance Using Decision Trees (NeurIPS 2017 · similarity 0.86)
- LEAPS: A discrete neural sampler via locally equivariant networks (ICML 2025 · similarity 0.85)
- Leveraging Importance Weights in Subset Selection (ICLR 2023 · similarity 0.85)
- Accelerating Neural Field Training via Soft Mining (CVPR 2024 · similarity 0.84)
- BARK: A Fully Bayesian Tree Kernel for Black-box Optimization (ICML 2025 · similarity 0.84)
- Importance Sampling for Nonlinear 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.