RankNet
Stub. Created as a placeholder — expand with vault-sourced content.
Definition
RankNet (Burges et al., 2005) is the foundational pairwise neural Learning to Rank approach: it models the probability that document A ranks above document B and trains with cross-entropy on those pairwise preferences. It is the conceptual precursor to LambdaRank and LambdaMART.
Vault references (existing coverage)
- Learning to Rank — listed as the foundational pairwise approach
TODO
- Expand with the lambda-gradient lineage RankNet → LambdaRank → LambdaMART.