Wyniki wyszukiwania - BazTech

Ograniczanie wyników

2 Bulletin of the Polish Academy of Sciences. Technical Sciences

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: IRLS

Sortuj według:

Ogranicz wyniki do:

Generalized ordered linear regression with regularization

Łęski J., Henzel N.

Bulletin of the Polish Academy of Sciences. Technical Sciences

2012

Vol. 60, nr 3

481-489

Linear regression analysis has become a fundamental tool in experimental sciences. We propose a new method for parameter estimation in linear models. The 'Generalized Ordered Linear Regression with Regularization' (GOLRR) uses various loss functions (including the o-insensitive ones), ordered weighted averaging of the residuals, and regularization. The algorithm consists in solving a sequence of weighted quadratic minimization problems where the weights used for the next iteration depend not only on the values but also on the order of the model residuals obtained for the current iteration. Such regression problem may be transformed into the iterative reweighted least squares scenario. The conjugate gradient algorithm is used to minimize the proposed criterion function. Finally, numerical examples are given to demonstrate the validity of the method proposed.

Iteratively reweighted least squares classifier and its l2- and l1-regularized Kernel versions

Łęski J.

Bulletin of the Polish Academy of Sciences. Technical Sciences

2010

Vol. 58, nr 1

171-182

This paper introduces a new classifier design method based on regularized iteratively reweighted least squares criterion function. The proposed method uses various approximations of misclassification error, including: linear, sigmoidal, Huber and logarithmic. Using the represented theorem a kernel version of classifier design method is introduced. The conjugate gradient algorithm is used to minimize the proposed criterion function. Furthermore, .1-regularized kernel version of the classifier is introduced. In this case, the gradient projection is used to optimize the criterion function. Finally, an extensive experimental analysis on 14 benchmark datasets is given to demonstrate the validity of the introduced methods.