Narrow Results

This paper presents a trust region subspace method for minimizing large-scale unconstrained problems. We choose a subspace that consists of some old directions which are invariable and some newest directions which are changed at each iteration. A restart technique is used when the old directions have little contribution. Numerical results are reported which indicate that the method is promising.

In many pattern recognition problems, it is desirable to reduce the number of input features by extracting important features related to the problems. By focusing on only the problem-relevant features, the dimension of features can be greatly reduced and thereby can result in a better generalization performance with less computational complexity. In this paper, we propose a feature extraction method for handling classification problems. The proposed algorithm is used to search for a set of linear combinations of the original features, whose mutual information with the output class can be maximized. The mutual information between the extracted features and the output class is calculated by using the probability density estimation based on the Parzen window method. A greedy algorithm using the gradient descent method is used to determine the new features. The computational load is proportional to the square of the number of samples. The proposed method was applied to several classification problems, which showed better or comparable performances than the conventional feature extraction methods.

In this paper, a stochastic model of plague is first studied by subspace identification. First, the discrete model of plague is obtained based on the classical model. The corresponding stochastic model is proposed for the existence of stochastic disturbances. Second, for the model, the parameter matrices and noise intensity are obtained. Finally, the simulations of the model show that the subspace identification is more precise than least square method.

The tissue characterization of coronary plaque is essential for a diagnosis of acute coronary syndromes. The radio frequency (RF) signals obtained by the intravascular ultrasound (IVUS) method are the representative data for tissue characterization of coronary plaque. RF signals are converted to luminosity values, and the cross sectional images of coronary artery can be observed, which is to evaluate the condition of acute coronary syndromes in real time. The images however are obscured by the attenuation of the ultrasound depending on the distance from the ultrasonic probe and by the ultrasound scattering due to red blood cells. Intravascular tissue characterization based on the analysis of RF signals, as an attempt to evaluate tissue structure quantitatively, have been proposed. Representative conventional methods use the analysis of the frequency characteristics of RF signals. However, the computing cost is high and the settings of the parameters are complicated and dificult. In this paper, tissue characterization of coronary plaque by subspace method is newly proposed. The proposed method, not only considers the neighborhood information around the target tissue, but also realizes a low computing cost and has easy parameter settings. The effectiveness of the proposed method has been verified by the application to the actual tissue characterization problem of the human coronary plaque.