Wyniki wyszukiwania - BazTech

1

Nonparametric methods of supervised classification

Jóźwik A.

Journal of Medical Informatics & Technologies

|

2013

|

Vol. 22

21--32

EN

Selected nonparametric methods of statistical pattern recognition are described. A part of them form modifications of the well known k-NN rule. To this group of the presented methods belong: a fuzzy k-NN rule, a pair-wise k-NN rule and a corrected k-NN rule. They can improve classification quality as compared with the standard k-NN rule. For the cases when these modifications would offer to large error rates an approach based on class areas determination is proposed. The idea of class areas can be also used for construction of the multistage classifier. A separate feature selection can be performed in each stage. The modifications of the k-NN rule and the methods based on determination class areas can be too slow in some applications, therefore algorithms for reference set reduction and condensation, for simple NN rule, are proposed. To construct fast classifiers it is worth to consider also a pair-wise linear classifiers. The presented idea can be used as in the case when the class pairs are linearly separable as well as in the contrary case.

2

Modyfikacja algorytmu Changa z wykorzystaniem metody znajdowania punktów najbliższych

Sierszeń A.

Automatyka / Akademia Górniczo-Hutnicza im. Stanisława Staszica w Krakowie

|

2007

|

T. 11, z. 3

193-202

PL

Modyfikacja algorytmu Changa polegająca na zastąpieniu oryginalnego sposobu wyznaczania odległości do najbliżej położonego punktu z tej samej klasy zmodyfikowanymi metodami znajdowania punktów najbliższych przynosi dużą akceleracje obliczeń. Przeprowadzone eksperymenty dowodzą, że zaproponowana metoda nie zmniejsza w sposób istotny jakości klasyfikacji.

EN

The modification of Chang's algorithm consisting in replacement of the original method of determining the distance to the nearest point from the same class with modified methods of finding the mutually nearest points causes a great acceleration of the computational phase. Results of experiments show that the presented method does not significantly decrease the quality of classification.

3

An example of computer aided decision-making system for recognition of respiration pathology

Jóźwik A., Sokołowska B., Budzińska K.

Journal of Medical Informatics & Technologies

|

2001

|

Vol. 2, Part. 1

MI41--MI48

EN

A main objective of the work was presentation of a new statistic approach to an analysis of respiration data. The breathing with intact and denervated diaphragm was compared. The respiration process was desciribed by three parameters: breathing frequency, tidal volume, and minute ventilation. Experimental data concerned a group of twelve anaesthetised cats. These data were analysed by a modification of the well-known k nearest neighbour rule (k-NN). It has been adopted from the statistical pattern recognition theory. The three ventilatory parameters were used to recognise whether we deal with the normal or the pathological case. Certain percentage of misclassifications must be taken into account. This misclassification rate is a measure how strong is the dependence between the ventilation parameters and preservation of the diaphragm innervation. The proposed method promises good differentiation of the two compared ways of respiration. It offers nearly five times smaller misclassification rate as compared with the standard k-NN rule.

4

A parallel net of (1-NN, K-NN) classifiers for optical inspection of surface defects in ferrites

Jóźwik A., Chmielewski L., Skłodowski M., Cudny W.

Machine Graphics and Vision

|

1998

|

Vol. 7, No. 1/2

99-112

EN

The challenging task of optical inspection of surface defects in ferrite cores has been successfully approached with a set of methods. In this paper the attention is paid to the k Nearest Neighbours classifier developed for the system. A parallel net of two-decision classifiers is presented. The combination of the 1-NN and k-NN rules reduces the training time. A great part of computations is restricted to the class overlap area. The classification quality is significantly improved if a separate feature selection for each of the component classifiers is done. A dramatic improvement of classification speed obtained by reference patterns sets reduction for component classifiers is vital, as in the considered task the classifier is used for recognition of pixels. The proposed modifications of the classifier are of general usefulness for pattern recognition. The presented quality inspection system can be applied to various defect detection tasks.