Bagging gradient-boosted trees for high precision, low variance ranking models

By Yasser Ganjisaffar, R. Caruana, Cristina Videira Lopes. 2011

214 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Optimization & Training

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Optimization & Training

Citations in the corpus

Pretrained Transformers for Text Ranking: BERT and Beyond (2020, background): “n the eve of the deep learning revolution, with the development of models based on tree ensembles [Burges, 2010].23 At that time, there was an emerging consensus that gradient-boosted decision trees [Ganjisaffar et al., 2011], in particular, represented the most effective solution to learning to ran”
End-to-End Answer Chunk Extraction and Ranking for Reading Comprehension (2016, methodology): “For the feature ranking-based system, we used jforest ranker (Ganjisaffar, Caruana, and Lopes 2011) with LambdaMART-RegressionTree algorithm and the ranking metric was NDCG@10.”
TREC Deep Learning Track: Reusable Test Collections in the Large Data Regime (2021, methodology): “[11] provide one demonstration of this approach being applied in information retrieval.”
Exploring demographic information in social media for product recommendation (2015, methodology): “As previous studies showed in [21], boosting mainly reduces the bias while bagging mainly reduces variance.”
A comparative study of machine learning classifiers for modeling travel mode choice (2017, methodology): “AC CE PT ED M AN US CR IP T boosting methods are primarily designed to minimize model bias and are therefore more prone to overfitting, 292 while RF and BAG conceptually aim to reduce model variance (Ganjisaffar et al., 2011).”
Query Driven Algorithm Selection in Early Stage Retrieval (2017, methodology): “In order to build a competitive “last stage” reference list, we train a Risk-Sensitive LambdaMartmodel using the JForests library [23, 52], and 687 queries from the 2009 Million Query Track (MQ2009) query set, which have shallow relevance judgments.”
Training Efficient Tree-Based Models for Document Ranking (2013, methodology): “[2] take this one step further and construct multiple ensembles, each built using a random bootstrap of the training data (i.”
Efficient and effective retrieval using selective pruning (2013, unknown): “We use the Jforests implementation [18].”
Effectiveness/efficiency tradeoffs for candidate generation in multi-stage retrieval architectures (2013, methodology): “Separately, we learned a LambdaMART model [5] using the open-source jforests implementation(4) [10].”