Lazy naive credal classifier

Corani, Giorgio; Zaffalon, Marco

doi:10.1145/1610555.1610560

Cited by 12 publications

(17 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although, in this case, IDM is still well-defined, we cannot let go sharp to zero in the optimization in (4). Therefore, the set of posteriors and the set of non-dominated classes will depend on the choice of .…”

Section: Restricting Idmmentioning

confidence: 99%

“…Instead, to compare credal classifiers we adopt two metrics which have been introduced in [4]. We refer to a classifier as accurate on a certain instance if its output includes the correct class, regardless how many classes it has returned; we refer to a classifier as determinate if its output contains only a single class.…”

Section: Comparing Credal Classifiersmentioning

confidence: 99%

“…Yet, there is no reason for linearly discounting the accuracy on the number of returned classes; an alternative non-parametric approach proposed in [4] removes this arbitrariness, being based on a rank test. The rank test is more robust than d-acc, as it does not encode any (arbitrary) functional form for discounting accuracy on the basis of the output size; yet, it uses less pieces of information than d-acc and can be therefore be less sensitive.…”

Section: Comparing Credal Classifiersmentioning

confidence: 99%

See 2 more Smart Citations

Restricting the IDM for Classification

Corani

Benavoli

2010

Communications in Computer and Information Science

Self Cite

View full text Add to dashboard Cite

Abstract. The naive credal classifier (NCC) extends naive Bayes classifier (NBC) to imprecise probabilities to robustly deal with the specification of the prior; NCC models a state of ignorance by using a set of priors, which is formalized by Walley's Imprecise Dirichlet Model (IDM). NCC has been shown to return more robust classification than NBC. However, there are particular situations (which we precisely characterize in the paper) under which the extreme densities included by the IDM force NCC to become very indeterminate, although NBC is able to issue accurately classifications. In this paper, we propose two approaches which overcome this issue, by restricting the set of priors of the IDM . We analyze both approaches theoretically and experimentally.

show abstract

Section: Restricting Idmmentioning

confidence: 99%

Section: Comparing Credal Classifiersmentioning

confidence: 99%

Section: Comparing Credal Classifiersmentioning

confidence: 99%

See 1 more Smart Citation

Restricting the IDM for Classification

Corani

Benavoli

2010

Communications in Computer and Information Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…In a recent survey on managing and mining uncertain data [2], the authors report only one SVM classifier dealing with classification with uncertain inputs. Note that the works of [16] about credal classifiers deal with uncertain inputs in the framework of imprecise probabilities where uncertainty is specified by a lower and upper bound probabilities for the observed value. In the probabilistic framework, most existing works concern classification with "fully certain observations" (which is not the topic of the paper).…”

Section: Related Workmentioning

confidence: 99%

Inference in possibilistic network classifiers under uncertain observations

Benferhat

Tabia

2012

Ann Math Artif Intell

View full text Add to dashboard Cite

Possibilistic networks, which are compact representations of possibility distributions, are powerful tools for representing and reasoning with uncertain and incomplete information in the framework of possibility theory. They are like Bayesian networks but lie on possibility theory to deal with uncertainty, imprecision and incompleteness. While classification is a very useful task in many real world applications, possibilistic network-based classification issues are not well investigated in general and possibilistic-based classification inference with uncertain observations in particular. In this paper, we address on one hand the theoretical foundations of inference in possibilistic classifiers under uncertain inputs and propose on the other hand a novel efficient algorithm for the inference in possibilistic networkbased classification under uncertain observations. We start by studying and analyzing the counterpart of Jeffrey's rule in the framework of possibility theory. After that, we address the validity of Markov-blanket criterion in the context of possibilistic networks used for classification with uncertain inputs purposes. Finally, we propose a novel algorithm suitable for possibilistic classifiers with uncertain observations without assuming any independence relations between observations. This algorithm guarantees the same results as if classification were performed using the possibilistic counterpart of Jeffrey's rule. Classification is achieved in polynomial time if the target variable is binary. The basic idea of our algorithm is to only search for 270 S. Benferhat, K. Tabia totally plausible class instances through a series of equivalent and polynomial transformations applied on the possibilistic classifier taking into account the uncertain observations.

show abstract

“…Consider T test samples and denote by b Y i the decision set obtained by the method and the selected decision rule for the ith test sample. The discounted accuracy is then (Tsoumakas and Vlahavas 2007;Corani and Zaffalon 2009) …”

Section: Using Lower Previsions To Choose Kmentioning

confidence: 99%

A K-nearest neighbours method based on imprecise probabilities

Destercke

2011

Soft Comput

View full text Add to dashboard Cite

K-nearest neighbours algorithms are among the most popular existing classification methods, due to their simplicity and good performances. Over the years, several extensions of the initial method have been proposed. In this paper, we propose a K-nearest neighbours approach that uses the theory of imprecise probabilities, and more specifically lower previsions. We show that the proposed approach has several assets: it can handle uncertain data in a very generic way, and decision rules developed within this theory allow us to deal with conflicting information between neighbours or with the absence of close neighbour to the instance to classify. We show that results of the basic k-NN and weighted k-NN methods can be retrieved by the proposed approach. We end with some experiments on the classical data sets.

show abstract

Lazy naive credal classifier

Cited by 12 publications

References 10 publications

Restricting the IDM for Classification

Restricting the IDM for Classification

Inference in possibilistic network classifiers under uncertain observations

A K-nearest neighbours method based on imprecise probabilities

Contact Info

Product

Resources

About