AI Paper Atlas

Research map / Learning theory and algorithms / Submodular and game algorithms

Core-sets for Fair and Diverse Data Summarization

Sepideh Mahabadi, Stojan Trajanovski

NeurIPS 2023 · Learning theory and algorithms › Submodular and game algorithms · approach: Streaming submodular algorithms · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
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=24Hurt
enforce fairness constraint (FDM) instead of DMSum-NN diversity maximization · Reddit% diversity loss: 9.65822% to 18.955% (uniform)Hurt
enforce fairness constraint (FDM) instead of DMMin-Pairwise diversity maximization · Reddit% diversity loss: 51.5744% -> 43.6022%Hurt
run plain DM without fairness constraintDM algorithms · Reddit, m=4, k=20color balanceHurt
increasing k_i (recency) instead of uniform k_iFDM · RedditSum-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 dataSum-Pairwise FDM · Reddit% diversity loss: 1.34723% -> 1.26935%No clear effect
apply FDM to union of core-sets instead of full dataSum-NN FDM · Reddit% diversity loss: 2.22301% to -3.02602%No clear effect
apply FDM to union of core-sets instead of full dataMin-Pairwise FDM · Reddit% diversity loss: 0.00%No clear effect
use core-sets for FDMSum-Pairwise FDM · Redditruntime speedup: 196.24199 -> 845.977729 times faster across configsImproved
use core-sets for FDMSum-NN FDM · Redditruntime speedup: 1769.7x at k=8 down to 23.7127x at k=60Improved
use our core-set instead of prior algorithmSum-Pairwise core-set · Redditunion core-set size: 12791 -> 40Improved
use our core-set instead of prior algorithmSum-Pairwise core-set · Reddit% diversity loss: 3.89448911685867% to 0.25503174848347%No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy