Leehter Yao

Nonlinear parameter estimation via the genetic algorithm

A modified genetic algorithm is used to solve the parameter identification problem for linear and nonlinear IIR digital filters. Under suitable hypotheses, the estimation error is shown to converge in probability to zero. The scheme is also applied to feedforward and recurrent neural networks. >

Sensor placement for on-orbit modal identification via a genetic algorithm

The selection and reproduction schemes of the genetic algorithm are modified, and a new operator called forced mutation is introduced. These changes are shown to improve the convergence of the algorithm and to lead to near-optimal sensor locations. Two practical examples are investigated: sensor placement for an early version of the space station and an individual space station photovoltaic array

Sensor placement for on-orbit modal identification of large space structure via a genetic algorithm

A variant of the genetic algorithm is used to place sensors optimally on a large space structure for the purpose of modal identification. The selection and reproduction schemes of the genetic algorithm are modified and a new operator called forced mutation is introduced. These changes are shown to improve the convergence of the algorithm and to lead to near-optimal sensor locations. Simulated results are also compared with previous results obtained by the effective independent method. The genetic algorithm based sensor configuration estimates the target mode response more accurately.<<ETX>>

Nonparametric learning of decision regions via the genetic algorithm

A method for nonparametric (distribution-free) learning of complex decision regions in n-dimensional pattern space is introduced. Arbitrary n-dimensional decision regions are approximated by the union of a finite number of basic shapes. The primary examples introduced in this paper are parallelepipeds and ellipsoids. By explicitly parameterizing these shapes, the decision region can be determined by estimating the parameters associated with each shape. A structural random search type algorithm called the genetic algorithm is applied to estimate these parameters. Two complex decision regions are examined in detail. One is linearly inseparable, nonconvex and disconnected. The other one is linearly inseparable, nonconvex and connected. The scheme is highly resilient to misclassification errors. The number of parameters to be estimated only grows linearly with the dimension of the pattern space for simple version of the scheme.

Identification of a nonlinear system modeled by a sparse Volterra series

An algorithm based on recursive approximation and estimation is proposed for the identification of nonlinear systems which can be modeled by a sparse Volterra series. The algorithm detects the terms of the Volterra series on which the output depends and estimates the associated Volterra kernels using a least squares criterion. The performance of the algorithm is primarily dependent on the number of nonzero Volterra kernels and not on their distribution in the whole series. The input sequence can be either i.i.d. or correlated. The algorithm can also be directly applied to the delay estimation of a sparse finite impulse response (FIR) filter.<<ETX>>

Track seeking hybrid fuzzy controller for the compact disc player

A hybrid track seeking fuzzy controller for an optical compact disc player is proposed. It is shown that the proposed hybrid fuzzy controller smoothes the applied voltage to the sled motor and improves the track seeking efficiency. The proposed hybrid fuzzy controller consists of three subsystems including parking time controller, driving force controller and parameter tuner. All three of subsystems are designed based on fuzzy inference. The hybrid fuzzy controller will on one hand drive the pickup head as fast as possible to the neighborhood of target track, it will on the other hand smoothly park the pickup head so that the fine seek controller can take over the control of pickup head.

Evolutionary design of fuzzy classifier with ellipsoidal decision regions

A fuzzy classifier using multiple ellipsoids approximating decision regions for classification is to be designed in this paper. An algorithm called Gustafson-Kessel algorithm (GKA) with an adaptive distance norm based on covariance matrices of prototype data points is adopted to learn the ellipsoids. GKA is able to adapt the distance norm to the underlying distribution of the prototype data points except that the sizes of ellipsoids need to be determined a priori. To overcome GKAs inability to determine appropriate size of ellipsoid, the genetic algorithm (GA) is applied to learn the size of ellipsoid. With GA combined with GKA, it is shown in this paper that the proposed method outperforms the benchmark algorithms as well as algorithms in the field.

An automatic hardness measuring method using Hough transform and fuzzy c-means algorithm

A novel automatic Vickers hardness measuring method is proposed. An algorithm called Hough fuzzy vertices detection algorithm (HFVDA) is proposed. In order to overcome the unavoidable affects of vertex detection due to surface contaminations or specimen texture, HFVDA transforms all the candidate pixels on the indentation edge lines into Hough space. Within Hough space, a weighted fuzzy c-means algorithm along with local maximum detection is proposed to find the transformed indentation edge lines. It will be shown that HFVDA is able to find the indentation vertices and calculate the hardness number with high accuracy for either specular-polished or rough-polished specimens.

Genetic algorithm based identification of nonlinear systems by sparse Volterra filters

In this paper, a sparse Volterra filter with parsimonious parametrization scheme is proposed. The sparse Volterra filter contains only the cross-products of input signals which contribute significantly to the system output. Based on the genetic algorithm, a scheme is proposed in this paper to automatically estimate the significant terms of cross-products of input signals. As the significant terms are detected, the associated Volterra kernels are estimated by the method of least square error. An operator called forced mutation is proposed to increase the rate of convergence of the genetic algorithm. Mathematical analysis is made to justify the effect of forced mutation.

An Evolutionary Fuzzy Classifier with Adaptive Ellipsoids

A fuzzy classifier using multiple ellipsoids approximating decision regions for classification is designed in this paper. We define a fuzzy rule to represent an ellipsoid decision region. An algorithm called Gustafson-Kessel Algorithm (GKA) with an adaptive distance norm based on covariance matrices of prototype data is adopted to learn the ellipsoids. GKA is able to adapt the distance norm to the prototype data except that the sizes of ellipsoids need to be determined a priori. To overcome GKAs inability to determine appropriate size of ellipsoid, the genetic algorithm (GA) is applied to learn the size of ellipsoid. With GA combined with GKA, the proposed method outperforms the benchmark algorithms as well as algorithms in the field.