Wyniki wyszukiwania - BazTech

1

Self-Adaptive Stable Mutation Based on Discrete Spectral Measure for Evolutionary Algorithms

Obuchowicz A., Prętki P.

Journal of Telecommunications and Information Technology

|

2011

|

nr 4

11-19

EN

In this paper, the concept of a multidimensional discrete spectral measure is introduced in the context of its application to the real-valued evolutionary algorithms. The notion of a discrete spectral measure makes it possible to uniquely define a class of multivariate heavy-tailed distributions, that have recently received substantial attention of the evolutionary optimization community. In particular, an adaptation procedure known from the distribution estimation algorithms (EDAs) is considered and the resulting estimated distribution is compared with the optimally selected referential distribution.

2

Learning stable mutation in (1, lambda)ES evolutionary strategy

Prętki P.

Prace Naukowe Politechniki Warszawskiej. Elektronika

|

2007

|

z. 160

233-240

EN

In this paper the concept of two-dimensional discrete spectral measure is introduced in the context of its application to real-valued evolutionary strategy (1, lambda)ES. The notion of discrete spectral measure makes it possible to uniquely define a class of multivariate heavy-tailed distributions, that have received more and more attention of evolutionary optimization community, recently. In particular, an adaptation procedure known from the class of estimation of distribution algorithms (EDAs) is proposed. The effectiveness of the evolutionary strategy is tested by means of a set popular benchmark functions.

3

Evolutionary bands for the expected response in nonlinear regression

Prętki P., Obuchowicz A.

Prace Naukowe Politechniki Warszawskiej. Elektronika

|

2006

|

z. 156

359-364

EN

In this paper evolutionary algorithms are applied to computation of confidence intervals for the expected response of nonlinear models. A simple phenotypic evolutionary algorithm was adapted to deal with nonlinear constraints and utilized to find the maximum and minimum value of a nonlinear model responses inside a confidence region. Moreover, the adequacy of the proposed approach is tested in a series of numerical simulations, and compared with the commonly applied linearization technique.

4

Designing neural-network-based fault detection systems with D-optimum experimental conditions

Witczak M., Prętki P.

Computer Assisted Mechanics and Engineering Sciences

|

2005

|

Vol. 12, No. 2-3

279-291

EN

The paper deals with an application of the theory of optimum experimental design to the problem of selecting the data set for developing neural models. Another objective is to show how to design a robust fault detection scheme with neural networks and how to increase its fault sensitivity by decreasing model uncertainty. It is also shown that the optimum design is independent of the parameters that enter linearly into the neural network. The final part of this paper shows a comprehensive simulation study regarding modelling and fault detection with the proposed approach. In particular, The DAMADICS benchmark problem is utilized to verify the performance and reliability of the proposed technique.

5

Phenotypic evolution with a mutation based on symmetric alpha-stable distributions

Obuchowicz A., Prętki P.

International Journal of Applied Mathematics and Computer Science

|

2004

|

Vol. 14, no 3

289-316

EN

Multidimensional Symmetric alpha-Stable (S alpha S) mutations are applied to phenotypic evolutionary algorithms. Such mutations are characterized by non-spherical symmetry for alpha<2 and the fact that the most probable distance of mutated points is not in a close neighborhood of the origin, but at a certain distance from it. It is the so-called surrounding effect (Obuchowicz, 2001b; 2003b). For alpha=2, the S alpha S mutation reduces to the Gaussian one, and in the case of alpha=1, the Cauchy mutation is obtained. The exploration and exploitation abilities of evolutionary algorithms, using S alpha S mutations for different alpha, are analyzed by a set of simulation experiments. The obtained results prove the important influence of the surrounding effect of symmetric alpha-stable mutations on both the abilities considered.