2 ms·
Useful for web 2.0: Given pairwise comparisons with some probability of error, how do you sort?
- amichail 19y ago"We study the problem of sorting in the presence of noise. While sorting linear orders is a classical well studied problem, the introduction of noise poses very interesting challenges. Noise has to be considered when ranking or sorting is applied in many real life scenarios. A natural example comes from sports. How do we rank a league of soccer teams based on the outcome of the games? It is natural to assume that there is a true underlying order of which team is better and that the games outcome represent noisy versions of the pairwise comparisons between teams. Note that in this problem it is impossible to aEURoere-sampleaEURX the order between a pair of teams. As a second example, consider experts comparing various items according to their importance where each pair of elements is compared by one expert. It is natural to assume that the experts opinions represent a noisy view of the actual order of significance. The question is then how to aggregate this information?"