Label ranking by learning pairwise preferences

By Eyke Hüllermeier, Johannes Fürnkranz, Weiwei Cheng, Klaus Brinker. Artificial Intelligence, 2008

589 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Management Science and Operations Research, Signal Processing

View Paper: https://www.semanticscholar.org/paper/76ac59ce9d8cd50fd2eebcb8cd7a0d01c3e43ea2

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 Management Science and Operations Research

In Signal Processing

Citations in the corpus

On label dependence and loss minimization in multi-label classification (2012, methodology): “Ranking by pairwise comparison (Hüllermeier et al. 2008; Fürnkranz et al. 2008) is a reduction method that transform the MLC problem to a quadratic number of binary problems, one for each pair of labels.”
Preference-Based Reinforcement Learning: A Preliminary Survey (2013, methodology): “This is then a sample for the relative utility function and used as training information for a label ranking algorithm [11] with φ(s) as features.”
Mining Multi-label Data (2010, methodology): “4: Data sets produced by the BR method Ranking by pairwise comparison (RPC) [32] transforms the multi-label dataset into M(M−1) 2 binary label datasets, one for each pair of labels (λ i , λ j ), 1 ≤ i j ≤ M.”
Preference-Based Policy Iteration: Leveraging Preference Learning for Reinforcement Learning (2011, background): “For training the label ranker we use LPC (cf. Section 3.1) with simple multi-layer perceptrons (as implemented in the Weka machine learning library [7] with its default parameters) as base classifiers.”
A Survey of Preference-Based Reinforcement Learning Methods (2017, background): “For deriving a ranking over all actions in a given state, the pairwise preference predictions of the trained classifiers Cij may be combined via voting or weighted voting (Hüllermeier et al., 2008), where each prediction issues a vote for its preferred action.”
Multilabel classification via calibrated label ranking (2008, methodology): “A subsequent transformation into the binary domain, such as for ranking by pairwise comparison , is a viable option, but other approaches for solving label ranking problems could also be used (Hüllermeier et al. 2008 contains a brief survey).”
Addressing imbalance in multilabel classification: Measures and random resampling algorithms (2015, methodology): “Among the transformation methods, the most popular are those based on the binarization of the MLD, such as Binary Relevance (BR) [11] and Ranking by Pairwise Comparison [12], and the LP [13] transformation, which produces a multiclass dataset from an MLD considering each labelset as class.”
Multilevel Language and Vision Integration for Text-to-Clip Retrieval (2018, methodology): “To train the retrieval model, we use a triplet-based retrieval loss, also called pairwise ranking loss (H¨ullermeier et al. 2008), which has shown good performance in ).”
A Literature Survey on Algorithms for Multi-label Learning (2010, methodology): “Ranking by Pairwise Comparison (RPC) Ranking by Pairwise Comparison (RPC) [24] transforms the multi-label dataset into ( k 2 ) binary label datasets, one for each pair of labels, (λi, λj), 1 ≤ i < j ≤ k.”
Active Preference-Based Learning of Reward Functions (2017, background): “Active preference-based learning has been successfully used in many domains [1, 6, 7, 14], but what makes applying it to learning reward functions difficult is the”
Fast Differentiable Sorting and Ranking (2020, methodology): “We consider the 21 datasets from (Hüllermeier et al., 2008; Cheng et al., 2009), which has both semi-synthetic data obtained from classification problems, and real biological measurements.”
A generative restricted Boltzmann machine based method for high-dimensional motion data modeling (2015, methodology): “To address this issue, we perform a label ranking process [31] to make the final decision using a set of pairwise classifiers.”
Label Distribution Learning (2013, background): “For example, if x represents a protein, y represents a cancer, then d y x should be the expression level of the protein x in the cancer y .”
Lift: Multi-Label Learning with Label-Specific Features (2011, methodology): “Calibrated label ranking (CLR) [14]: This is a secondorder approach which transforms the multi-label learning problem into a label ranking problem, where pairwise comparison [21] is employed to train”