Ranking Support Vector Machine with Kernel Approximation

Table 5

Performance comparison on MQ2007 data set.

	NDCG@1	NDCG@3	MeanNDCG	P@1	P@3	MAP
RankSVM-Struct [9]	0.4096	0.4063	0.4966	0.4746	0.4315	0.4645
ListNet [19]	0.4002	0.4091	0.4988	0.4640	0.4334	0.4652
AdaRank-MAP [20]	0.3821	0.3984	0.4891	0.4392	0.4230	0.4577
AdaRank-NDCG [20]	0.3876	0.4044	0.4914	0.4475	0.4305	0.4602
RankBoost [18]	0.4134	0.4072	0.5003	0.4823	0.4348	0.4662
LambdaMART [25]	0.4147	0.4119	0.5011	—	—	0.4660
BL-MART [25]	0.4200	0.4224	0.5093	—	—	0.4730
CRR [26]	—	—	0.5000	—	—	0.4660
RankSVM-Primal [8]	0.4109	0.4063	0.4973	0.4747	0.4317	0.4655
RankNystöm	0.4242	0.4138	0.5036	0.4888	0.4394	0.4695
RankRandomFourier	0.4224	0.4136	0.5036	0.4871	0.4386	0.4698