Learning "Learning to Rank"
summary
Covers how to turn query–document pairs into numeric feature vectors (IDF, term-in-title, link counts), the need for human relevance labels (0–4), and comparisons of pointwise (linear regression), pairwise (RankNet, LambdaRank) and listwise (LambdaMART) approaches using normalized discounted cumulative gain as the evaluation metric. Demonstrates Python examples and tree visualization with XT boost (tuning eta and max_depth), reports experiments on the Microsoft learning to rank web OK dataset (~2.6M rows), and warns about overfitting and feature engineering.
AI-written summary · about
chapters (7)
- 0:00Start
- 6:44Pointwise Ranking
- 13:07Pairwise Ranking
- 13:46RankNet
- 15:10Lambda Rank
- 15:47Listwise Ranking
- 16:10LambdaMART
related talks
29mBuilding Learning to Rank models for search using LLMs
38mLearning to Rank: where search meets machine learning #bbuzz
55mLearning to Rank, a Little
47mLearning to rank search results
43mLearning to Rank for Reddit Search: A Project Retro
29mLearning to rank with the Transformer
43m[Machine Learning] Creating more relevant search results with Learn to Rank
25mBetter search relevance using Learning to Rank at mobile.de