Rejection strategies and confidence measures for a k-NN classifier in an OCR task

Arlandis, Joaquim; Pérez-Cortés, Juan-Carlos; Cano, Jesus

doi:10.1109/icpr.2002.1044806

Cited by 24 publications

(28 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A revision of the expression to compute the a posteriori probability based on [15] and [2] applied to k-NN is used. In this case, the estimation involves computing the nearest neighbor prototype for each class and normalizing the inverse of the distance.…”

Section: Verification System Schemesmentioning

confidence: 99%

“…The normalized sigmoids of the distances to the discriminant functions are taken as the estimates for the posterior probabilities. On the other hand, Arlandis et al [2] proposed an interesting method based on the k-NN, focused only on the distances to the k-NN prototypes, assigning a zero probability to the prototypes outside that neighborhood. Nevertheless, in most cases, the zero probability is not a realistic situation.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Confidence voting method ensemble applied to off-line signature verification

Rico-Juan

Iñesta

2012

Pattern Anal Applic

View full text Add to dashboard Cite

In this paper, a new approximation to off-line signature verification is proposed based on two-class classifiers using an expert decisions ensemble. Different methods to extract sets of local and a global features from the target sample are detailed. Also a normalization by confidence voting method is used in order to decrease the final equal error rate (EER). Each set of features is processed by a single expert, and on the other approach proposed, the decisions of the individual classifiers are combined using weighted votes. Experimental results are given using a subcorpus of the large MCYT signature database for random and skilled forgeries. The results show that the weighted combination outperforms the individual classifiers significantly. The best EER obtained were 6.3 % in the case of skilled forgeries and 2.31 % in the case of random forgeries.

show abstract

Section: Verification System Schemesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Confidence voting method ensemble applied to off-line signature verification

Rico-Juan

Iñesta

2012

Pattern Anal Applic

View full text Add to dashboard Cite

show abstract

“…In such a case, the option of rejection to classify a pattern should be applied. The reject option is a subject of many works (Cordella et al 1995, Ha 1996a, Baram 1998, Fumera et al 2000, Duda et al 2001, Sansone et al 2001, Arlandis et al 2002, SantosPereira and Pires 2005. The principal approach in analysing the trade-off between a potential error and rejection is conventionally formulated in terms of posterior probabilities of different classes: if the posterior probability of the best class label is high, then the label is attached to the pattern and reported; if not, the classification is not performed (reject option); see, for example, Battiti and Colla (1994).…”

Section: Introductionmentioning

confidence: 99%

Choosing a discernibility measure for reject-option of individual and multiple classifiers

Voulgaris

Mirkin

2010

International Journal of General Systems

View full text Add to dashboard Cite

A novel method for evaluating the reliability of a classifier on a pattern is proposed based on the discernibility of a pattern's class against other classes from the pattern. Three measures of discernibility are proposed and experimentally compared with each other and with more conventional techniques based on the classification scores for class labels. The classification accuracy can be significantly enhanced through discernibility measures using the most reliable -'elite' -patterns. It can be further boosted by forming an amalgamation of the elites of different classifiers. Improved performance is achieved at the price of rejecting many patterns. There are situations in which this price is worth paying -when the non-reliable predictions, however good, lead to the need for the manual testing of very cumbersome and complex technical devices or in diagnostics of human terminal diseases. Contrary to conventional techniques for estimating reliability, the proposed measures are applicable to small datasets as well as to datasets with complex class structures on which conventional classifiers show low accuracy rates.

show abstract

“…The condences of the single classiers are based on the Euclidean distance to the training set samples. We use a modication of the expression to compute a posteriory probability based on [3] and proposed in [4]. The estimation isP (ω i |x) where ω i is the i-th class and x is a new example to classify.…”

Section: Classication Schemesmentioning

confidence: 99%

“…There are some works which use condence methods based in a posteriory probabilities (according to the Bayes theory) in classication, as in [3], where several formulas are proposed to estimate this probability, and in [4], where the authors propose a method based on the k-Nearest Neighbour rule.…”

Section: Introductionmentioning

confidence: 99%

Normalisation of Confidence Voting Methods Applied to a Fast Handwritten OCR Classification

Rico-Juan

Iñesta

2007

Advances in Soft Computing

View full text Add to dashboard Cite

Summary. In this work, a normalisation of the weights utilized for combining classiers decisions based on similarity Euclidean distance is presented. This normalisation is used by the condence voting methods to decrease the nal error rate in an OCR task. Dierent features from the characters are extracted. Each set of features is processed by a single classier and then the decisions of the individual classiers are combined using weighted votes, using dierent techniques. The error rates obtained are as good or slightly better than those obtained using a Freeman chain codes as contour representation and the string edit distance as similarity measure, but the complexity and classication time decrease dramatically.

show abstract

Rejection strategies and confidence measures for a k-NN classifier in an OCR task

Cited by 24 publications

References 4 publications

Confidence voting method ensemble applied to off-line signature verification

Confidence voting method ensemble applied to off-line signature verification

Choosing a discernibility measure for reject-option of individual and multiple classifiers

Normalisation of Confidence Voting Methods Applied to a Fast Handwritten OCR Classification

Contact Info

Product

Resources

About