Multi-label Classification Using Ensembles of Pruned Sets

By Jesse Read, Bernhard Pfahringer, G. Holmes. 2008 Eighth IEEE International Conference on Data Mining, 2008

475 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/6c2e32948af0134f0c3d85e8c262345568912208

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 Information Systems

Citations in the corpus

An extensive experimental comparison of methods for multi-label learning (2012, methodology): “Decision trees : Clare et al. [11] adapted the C4.5 algorithm for multi-label data (ML-C4.5) by modifying the formula for calculating entropy.”
Lift: Multi-Label Learning with Label-Specific Features (2011, methodology): “…[43]: This is a high-order approach which transforms the multi-label learning problem into an ensemble of multi-class classification problems, where each multi-class classification problem is generated by applying the label powerset techniques [37], [59] on a randomly chosen k-labelset in Y.”
Label Distribution Learning (2013, background): “For example, t h MLL problem could be transformed into binary classification problems [14], [25], a label ranking problem [15], [9], or an ensemble learning problem [26], [17].”
Multi-label learning by exploiting label dependency (2010, background): “• High-order approaches: The task of multi-label learning is tackled by considering the high-order relations between labels, such as the full-order style of imposing all other labels’ influences on each label in an indirect manner [3, 10, 11, 25], or the random style of combining an ensemble of clas”
Learning to diagnose from scratch by exploiting dependencies among labels (2017, methodology): “segmentation tasks such as in Ronneberger et al. (2015); Milletari et al.”
A Tutorial on Multilabel Learning (2015, methodology): “Nevertheless, its build time increases by a factor of approximately 10 each time k is doubled [Read et al. 2008].”
Multi-target regression via input space expansion: treating targets as inputs (2012, background): “The same holds for approaches that consider sets of labels as diﬀerent values of a single-label multi-class task [26,32].”
Multi-label classification with a reject option (2013, background): “, [23]), it is widely used due to its limited computational cost.”
MLSMOTE: Approaching imbalanced multilabel learning through synthetic instance generation (2015, methodology): “There are many proposals in this field, from MLC trees like ML-TREE [20] or a multilabel kNN called ML-kNN [21] to multilabel neural networks [22,23] and SVMs [24].”
On the Stratification of Multi-label Data (2011, methodology): “For example, the stratified sampling approaches discussed in this paper could be used for reduced error pruning of multi-label decision trees [24], for down-sampling without replacement in the ensembles of pruned sets approach [17] and for deciding when to stop the training of a multi-label neural n”
Scalable and efficient multi-label classification for evolving data streams (2012, methodology): “Of all the multi-label classifiers that scale well with the size of the data we prefer to use the Pruned Sets method (PS) (Read et al. 2008).”
A Literature Survey on Algorithms for Multi-label Learning (2010, background): “in [42] calls LCard as a measure of ”multi-labelledness”.”
Multilabel Classification with Principal Label Space Transformation (2012, methodology): “For instance, the pruned label power set approach (Read, Pfahringer, & Holmes, 2008 ), which is a variant of the usual label power set, considers only vertices occupied by enough examples during training; topic modeling (Law et al, 2010 ) groups the occupied vertices as clusters; and kernel depend”
Classifier chains for multi-label classification (2009, methodology): “The process for generating such data is outlined in Read et al. (2009a) (since we are only measuring running time, we do not detail the data composition here).”