Bayes Optimal Multilabel Classification via Probabilistic Classifier Chains

By Krzysztof Dembczynski, Weiwei Cheng, Eyke Hüllermeier. 2010

506 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods, Statistical Learning Theory

How researchers see this paper

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

In Probabilistic & Bayesian Methods

In Statistical Learning Theory

Citations in the corpus

Elucidating image-to-set prediction: An analysis of models, losses and datasets (2019, neutral): “In order to explicitly exploit label co-occurrences, researchers have resorted to modeling the joint probability distribution of labels [59] or decomposing the joint distribution into conditionals [11, 61, 45, 28, 32, 41, 46].”
DeepCoder: Learning to Write Programs (2016, background): “Dembczynski et al. (2010) showed that in multi-label classification under a so-called Rank loss, it is Bayes optimal to rank the labels according to their marginal probabilities.”
FIRST (2018, unknown): “Bayes estimators are probabilistic classifiers [9], the same type of the well-known Naïve Bayes classifier.”
Large-Scale Multi-label Text Classification - Revisiting Neural Networks (2013, background): “In such cases, it is also intractable to employ methods that minimize ranking errors among labels [6, 31] or that learn joint probability distributions of labels [10, 3].”
Probabilistic Multi-Label Classification with Sparse Feature Learning (2013, unknown): “…2006; Fuernkranz et al., 2008], pairwise label dependency methods [Zhu et al., 2005; Petterson and Caetano, 2011], probabilistic classifier chains [Dembczynski et al., 2010], large-margin methods [Guo and Schuurmans , 2011; Godbole and Sarawagi, 2004; Hariharan et al., 2010], and probabilistic…”
On the Consistency of Multi-Label Learning (2011, unknown): “It is noteworthy that, after we published our preliminary work (Gao and Zhou, 2011), Dembczyński et al. (2012a) proved that the following surrogate losses are consistent w.”
Efficient Multi-label Classification with Many Labels (2013, background): “In recent years, many approaches have been proposed to further improve classification performance by incorporating the label correlations (Dembczynski et al., 2010; Hariharan et al., 2010) or exploiting the label hierarchy (Bi & Kwok, 2011).”
Improving Pairwise Ranking for Multi-label Image Classification (2017, background): “The Bayes consistency is an important property for a loss function to achieve the right objective [5, 24].”
Augment and Reduce: Stochastic Inference for Large Categorical Distributions (2018, methodology): “They might replace the softmax transformation with a hierarchical or stickbreaking model (Kurzynski, 1988; Morin & Bengio, 2005; Tsoumakas et al., 2008; Beygelzimer et al., 2009; Dembczyński et al., 2010; Khan et al., 2012).”
Remarks on multi-output Gaussian process regression (2018, background): “Besides, some theoretical insights [68, 69, 70] pointed out that the consideration of additional features to transform the training data helps SST and ERC reduce the model bias at the cost of increasing the model variance.”
Maximizing Subset Accuracy with Recurrent Neural Networks in Multi-label Classification (2017, methodology): “[3] have proposed probabilistic classifier chain (PCC) which decomposes the joint probability into L conditional probabilities:”
General Multi-label Image Classification with Transformers (2020, background): “(1) Conditional Prediction: The first type, autoregressive models [12, 41, 50, 37] estimate the true joint probability of output labels given the input by using the chain rule, predicting one label at a time.”
A Literature Survey on Algorithms for Multi-label Learning (2010, methodology): “Notice that the analysis in this section is based on the experiments and results described in [52, 20, 69, 14, 68] with a few additional alignment experiments conducted.”
Inverse Cooking: Recipe Generation From Food Images (2018, background): “Significant effort has been devoted in the literature to leverage deep neural networks for multi-label classification, by designing models [49, 8, 56, 37, 53] and studying loss functions [12] well suited for this task.”