Wyniki wyszukiwania - Biblioteka Nauki

1

Some Decompositions of Matrix Variances

100%

Léka Z. , Petz D.

|

tom Vol. 33, Fasc. 2

191--199

EN

When D is a density matrix and A1, A2 are self-adjoint operators, then the standard variance is a 2 × 2 matrix: VarD(A1, A2)i,j := TrDAiAj − (TrDAi)(TrDAj) (1 ≤ i, j ≤ 2). The main result in this work is that there are projections Pk such that D = Σk λk Pk with 0 < λk and Σk λk = 1 and VarD (A1, A2) = Σk λk VarPk (A1, A2). In a previous paper only the A1 = A2 case was included and the relevance is motivated by the paper [8].

2

Estimation of covariance parameters for GNSS/leveling geoid data by Leave-One-Out validation

72%

Jarmołowski W.

|

tom nr 16(4)

291--307

EN

The article describes the estimation of covariance parameters in Least Squares Collocation (LSC) by Leave-One-Out (LOO) validation, which is often considered as a kind of cross validation (CV). Two examples of GNSS/leveling (GNSS/lev) geoid data, characterized by different area extent and resolution are applied in the numerical test. A special attention is focused on the noise, which is not correlated in this case. The noise variance is set to be homogeneous for all points. Two parameters in three covariance models are analyzed via LOO, together with a priori noise standard deviation, which is a third parameter. The LOO validation finds individual parameters for different applied functions i.e. different correlation lengths and a priori noise standard deviations. Diverse standard deviations of a priori noise found for individual datasets illustrate a relevance of applying LOO in LSC. Two examples of data representing different spatial resolutions require individual noise covariance matrices to obtain optimal LSC results in terms of RMS in LOO validation. The computation of appropriate a priori noise variance is however difficult via typical covariance function fitting, especially in the case of sparse GNSS/leveling geoid data. Therefore LOO validation may be helpful in describing how the a priori noise parameter may affect LSC result and a posteriori error.

3

Widmo iloczynowe, widmo poliharmoniczne, korelacja i koherencja jako źródło danych wejściowych dla sztucznych sieci neuronowych

72%

Łazarz B. , Czech P.

Zeszyty Naukowe. Transport / Politechnika Śląska

|

2003

|

tom z. 50

159-164

PL

W opracowaniu przedstawiono wyniki eksperymentu mającego na celu znalezienie miar określających stopień uszkodzenia kół zębatych. Poszukiwano miar, które następnie posłużyłyby jako dane wejściowe w badaniach nad zastosowaniem sztucznych sieci neuronowych w diagnostyce przekładni zębatych.

EN

The work presents results of an experiment carried out in order to define simple measures which fully describe the degree of tooth root cracking. The key feature of the measures sought was the ability of using them as input data for a research in the application of Artificial Neural Networks in the tooth gear diagnostics.

4

Modelowanie kowariancji kursów walutowych z zastosowaniem cen minimalnych i maksymalnych

72%

Fiszeder P.

|

nr 3/2018 (76)

37-49

EN

The article presents a proposal for exchange rate modelling by means of minimum and maximum prices that enables a better description of dependencies in the foreign exchange market. Forecasts of return rate covariance based on the proposed multiple-equation GARCH model are more accurate than those produced solely on the basis of closing prices. Multiple-equation GARCH models are among the most popular models describing financial time series. The proposed model does not require additional data because daily minimum and maximum prices are generally available together with closing prices, which is important from the point of view of the application of the model in the Forex market.

PL

W artykule przedstawiono propozycję modelowania kursów walutowych z zastosowaniem cen minimalnych i maksymalnych, która prowadzi do lepszego opisu zależności na rynku walutowym. Konstruowane na podstawie zaproponowanego wielorównaniowego modelu GARCH prognozy kowariancji stóp zwrotu są trafniejsze niż prognozy konstruowane na podstawie wyłącznie cen zamknięcia. Wielorównaniowe modele GARCH należą do najbardziej popularnych modeli opisujących finansowe szeregi czasowe. Przedstawiona propozycja modelu nie wymaga pozyskania dodatkowych danych, ponieważ dzienne ceny minimalne i maksymalne są na ogół dostępne równolegle z cenami zamknięcia, co jest ważne z punktu widzenia aplikacji modelu na rynku Forex.

5

On a characterization of a power distribution

58%

Pawlas P. , Szynal D.

|

tom Vol. 48, nr 1

100--106

EN

A measure of dependence called pseudo-covariance and related to covariance was proposed by Pawlas and Szynal [4]. It was used, among other things, in a characterization of a power distribution. Here we generalized that result.

6

On a characterization of exponential, Pearson and Pareto distributions via covariance and pseudo-covariance

58%

Pawlas P. , Szynal D.

|

2020

|

tom Vol. 53, nr 1

285--291

EN

Properties of linear regression of order statistics and their functions are usually utilized for the characterization of distributions. In this paper, based on such statistics, the concept of Pearson covariance and the pseudo-covariance measure of dependence is used to characterize the exponential, Pearson and Pareto distributions.

7

Speech nonfluency detection and classification based on linear prediction coefficients and neural networks

51%

Kobus A. , Kuniszyk-Jóźkowiak W. , Smołka E. , Codello I.

|

tom Vol. 15

135--143

EN

The goal of the paper is to present a speech nonfluency detection method based on linear prediction coefficients obtained by using the covariance method. The application “Dabar” was created for research. It implements three different methods of LP with the ability to send coefficients computed by them into the input of Kohonen networks. Neural networks were used to classify utterances in categories of fluent and nonfluent. The first one was Kohonen network (SOM), used to reduce LP coefficients representation of each window, which were used as input data to SOM input layer, to a vector of winning neurons of SOM output layer. Radial Basis Function (RBF) networks, linear networks and Multi-Layer Perceptrons were used as classifiers. The research was based on 55 fluent samples and 54 samples with blockades on plosives (p, b, d, t, k, g). The examination was finished with the outcome of 76% classifying.