Research map / Learning theory and algorithms / Submodular and game algorithms
Core-sets for Fair and Diverse Data Summarization
Sepideh Mahabadi, Stojan Trajanovski
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 |
|---|---|---|---|
| enforce fairness constraint (FDM) instead of plain DM, uniform k_i=[2,2,2,2] | Sum-Pairwise diversity maximization · Reddit | % diversity loss: 1.21847% -> 0.193663% loss from k=8 to k=24 | Hurt |
| enforce fairness constraint (FDM) instead of DM | Sum-NN diversity maximization · Reddit | % diversity loss: 9.65822% to 18.955% (uniform) | Hurt |
| enforce fairness constraint (FDM) instead of DM | Min-Pairwise diversity maximization · Reddit | % diversity loss: 51.5744% -> 43.6022% | Hurt |
| run plain DM without fairness constraint | DM algorithms · Reddit, m=4, k=20 | color balance | Hurt |
| increasing k_i (recency) instead of uniform k_i | FDM · Reddit | Sum-Pairwise % diversity loss: 1.21847% -> 0.416084% (k=8 vs k=20 comparison approx) | Improved |
| apply FDM to union of core-sets instead of full data | Sum-Pairwise FDM · Reddit | % diversity loss: 1.34723% -> 1.26935% | No clear effect |
| apply FDM to union of core-sets instead of full data | Sum-NN FDM · Reddit | % diversity loss: 2.22301% to -3.02602% | No clear effect |
| apply FDM to union of core-sets instead of full data | Min-Pairwise FDM · Reddit | % diversity loss: 0.00% | No clear effect |
| use core-sets for FDM | Sum-Pairwise FDM · Reddit | runtime speedup: 196.24199 -> 845.977729 times faster across configs | Improved |
| use core-sets for FDM | Sum-NN FDM · Reddit | runtime speedup: 1769.7x at k=8 down to 23.7127x at k=60 | Improved |
| use our core-set instead of prior algorithm | Sum-Pairwise core-set · Reddit | union core-set size: 12791 -> 40 | Improved |
| use our core-set instead of prior algorithm | Sum-Pairwise core-set · Reddit | % diversity loss: 3.89448911685867% to 0.25503174848347% | No clear effect |
Closest papers by meaning
- Individually Fair Diversity Maximization (NeurIPS 2025 · similarity 0.89)
- Fair and Diverse DPP-Based Data Summarization (ICML 2018 · similarity 0.87)
- Fair k-Center Clustering for Data Summarization (ICML 2019 · similarity 0.87)
- Improved Streaming Algorithm for Fair k-Center Clustering (AAAI 2026 · similarity 0.87)
- New Coresets for Fair Clustering (NeurIPS 2026 · similarity 0.87)
- How to Solve Fair k-Center in Massive Data Models (ICML 2020 · similarity 0.86)
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.
