Research map / Learning theory and algorithms / Prediction and decision losses
Finding Most Influential Sets
Lucas Darius Konrad, Nikolas Kuschnig
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 exact Dinkelbach MIS instead of greedy selection | MIS selection · Optimization exactness (n<=50, k<=3) | optimality vs enumeration: always matches the enumerated optimum, while greedy sometimes returns suboptimal sets | Improved |
| warm-start Dinkelbach with adversarial initial eta values (ablation) | Dinkelbach MIS · n=10^8, k=10^7 | iterations: from five to six | No clear effect |
| use exact MIS rather than greedy on Mongolia microcredit RCT (decreasing direction) | MIS selection · Mongolia microcredit RCT | influence gap | Improved |
| use exact MIS rather than greedy on benchmark regression datasets | MIS selection · Benchmark regressions | influence value | No clear effect |
| increase query similarity from low to high in linearized STS regression | Linearized text embedding regression · STS benchmark | sensitivity to 5,000-MIS removal: Outside 0.04 -> 0.07/0.02; Romney 0.81 -> 1.43/0.41 | Hurt |
| use residualization with cross-fitting on partial linear simulation | MIS estimator · Residualized simulation, n in {1000, 5000, 10000} | influence value vs oracle and set recovery | Mixed |
| compare Dinkelbach MIS with greedy and influence approximations for runtime | MIS computation · Synthetic grid of (n, k) | runtime: often by an order of magnitude | Improved |
Closest papers by meaning
- Testing Most Influential Sets (ICLR 2026 · similarity 0.87)
- Most Influential Subset Selection: Challenges, Promises, and Beyond (NeurIPS 2024 · similarity 0.86)
- Learning Set Functions with Implicit Differentiation (AAAI 2025 · similarity 0.83)
- Sampling for Beyond-Worst-Case Online Ranking (AAAI 2024 · similarity 0.81)
- Ranking with Multiple Oracles: From Weak to Strong Stochastic Transitivity (ICML 2025 · similarity 0.81)
- Enhancing Neural Subset Selection: Integrating Background Information into Set Representations (ICLR 2024 · 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.
