Sampling pairs directly with auxiliary information for higher inclusion probabilities on informative pairs yields near-full pairwise loss performance at reduced computational cost.
ℓ(Zi, Zj) ¯πi,j(WN) − ℓ(Zk, Zl) ¯πk,l(WN) 2 WN # × ¯πi,j(WN)¯πk,l(WN)−¯π(i,j),(k,l)(WN) .(44) For all(i, j),(k, l)s.t.i < jandk < l, we have E
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Doing well with less! On Sampling Techniques for Empirical Pairwise Loss Estimation/Minimization
Sampling pairs directly with auxiliary information for higher inclusion probabilities on informative pairs yields near-full pairwise loss performance at reduced computational cost.