Kernelized Sorting (Poster ID: W48) Novi Quadrianto, Le Song, and Alex J. Smola Problem : Given two sets of observations X = {(x i )} m i=1 and Y = {(y i )} m i=1 , sort y i s according to x i s without a cross domain comparison. Solution : Maximizing dependence by permuting y i s π * = argmax π∈Π trace(HKH π Lπ > ); K ij = k (x i ,x j ),L ij = l (y i ,y j ) Applications : Image Layout Multilingual Document Matching English German NIPS is great! NIPS ist klasse! Our paper is cool! Unsere Veröffentlichung ist cool! Image Matching
Kernelized Sorting (Poster ID: W48)users.cecs.anu.edu.au/~nquadrianto/nips09_spotlight.pdf · Kernelized Sorting (Poster ID: W48) Novi Quadrianto, Le Song, and Alex J. Smola Problem
Kernelized Sorting (Poster ID: W48)Novi Quadrianto, Le Song, and
Alex J. SmolaProblem : Given two sets of observations X =
{(xi)}mi=1 and Y ={(yi)}mi=1, sort yis according to xis without a
cross domain comparison.
Solution : Maximizing dependence by permuting yisπ∗ = argmax
![[10] sorting](https://img.pdfslide.tips/doc/110x75/55b27a9bbb61eb95158b475e/10-sorting.jpg)
