Project Description
ListNetRanker is a listwise ranker based on machine learning. Given documents and query's feature set, ListNetRanker will rank these documents by ranking score.


ListNetRanker's core ranking algorithm is based on ListNet(Z. Cao, T. Qin, T. Liu, et al. ICML 2007). In training phrase, ListNetRanker generates training corpus's ranking probability distribution and In predict phrase, it uses KL distance as loss function to measure how similairy between training corpus's ranking probability distribution and test case's ranking probability distribution.

Last edited Aug 14, 2012 at 10:07 AM by monkeyfu, version 3