In this study, our purpose was to improve the performance of our mass detection system by using a new dual system approach which combines a computer-added detection (CAD) system optimized with "average" masses with another CAD system optimized with "subtle" masses. The two single CAD systems have similar image processing steps, which include prescreening, object segmentation, morphological and texture feature extraction, and false positive (FP) reduction by rule-based and linear discriminant analysis (LDA) classifiers. A feed-forward backpropagation artificial neural network was trained to merge the scores from the LDA classifiers in the two single CAD systems and differentiate true masses from normal tissue. For an unknown test mammogram, the two single CAD systems are applied to the image in parallel to detect suspicious objects. A total of three data sets were used for training and testing the systems. The first data set of 230 current mammograms, referred to as the average mass set, was collected from 115 patients. We also collected 264 mammograms, referred to as the subtle mass set, which were one to two years prior to the current exam from these patients. Both the average and the subtle mass sets were partitioned into two independent data sets in a cross validation training and testing scheme. A third data set containing 65 cases with 260 normal mammograms was used to estimate the FP marker rates during testing. When the single CAD system trained on the average mass set was applied to the test set with average masses, the FP marker rates were 2.2, 1.8, and 1.5 per image at the case-based sensitivities of 90%, 85%, and 80%, respectively. With the dual CAD system, the FP marker rates were reduced to 1.2, 0.9, and 0.7 per image, respectively, at the same case-based sensitivities. Statistically significant (p < 0.05) improvements on the free response receiver operating characteristic curves were observed when the dual system and the single system were compared using the test sets with either average masses or subtle masses.

Dual system approach to computerâ aided detection of breast masses on mammograms

Chaotic time series prediction with residual analysis method using hybrid Elman-NARX neural networks

http://plaza.ufl.edu/yiz21cn/refer/A%20comparative%20study%20of%20autoregressive%20neural%20network%20hybrids.pdf

2005 Special Issue: A comparative study of autoregressive neural network hybrids

Software reliability is an important factor for quantitatively characterizing software quality and estimating the duration of software testing period. Traditional parametric software reliability growth models (SRGMs) such as nonhomogeneous Poisson process (NHPP) models have been successfully utilized in practical software reliability engineering. However, no single such parametric model can obtain accurate prediction for all cases. In addition to the parametric models, non-parametric models like neural network have shown to be effective alternative techniques for software reliability prediction. In this paper, we propose a non-parametric software reliability prediction system based on neural network ensembles. The effects of system architecture on the performance are investigated. The comparative studies between the proposed system with the single neural network based system and three parametric NHPP models are carried out. The experimental results demonstrate that the system predictability can be significantly improved by combing multiple neural networks.

Predicting software reliability with neural network ensembles

Enhanced ensemble structures using wavelet neural networks applied to short-term load forecasting

We describe the use of ensemble methods to build proper models for time series prediction. Our approach extends the classical ensemble methods for neural networks by using several different model architectures. We further suggest an iterated prediction procedure to select the final ensemble members.

Time series prediction with ensemble models

Time series prediction with ensemble models applied to the CATS benchmark

We present a new method for detecting dependencies in the stock market. In order to find hidden correlations in the daily returns, we build cross prediction models and use the normalized modeling error as a generalized correlation measure that extends the concept of the classical correlation matrix.

/pdf/detecting-correlation-in-stock-market-3cela02zf0.pdf

Detecting correlation in stock market

We show how to train Discrete Time Cellular Neural Networks (DT-CNN) successfully by backpropagation to perform pattern recognition on a data set of handwritten digits. By using concepts and techniques from Machine Learning, we can outperform Support Vector Machines (SVM) on this problem.

Stochastic Gradient Descent Training of Ensembles of DT-CNN Classifiers for Digit Recognition

We introduce a software toolbox for building ensembles of computer models based on measured time series The toolbox is using statistical learning techniques for training of individual linear and nonlinear models as well as the construction of ensembles of heterogenous models types Several well performing model types among which are ridge regression, k-nearest neighbor models and neural networks have been implemented Ensembles of heterogenous models typically give better performance then any other class Additionally, methods for validation and model assessment are given as well as adaptor classes perform a transparent feature selection or the training random subsets of all input variables The toolbox is implemented in Matlab and C++

J.D. Wichard

Papers

Time series prediction with ensemble models

Time series prediction with ensemble models applied to the CATS benchmark

Detecting correlation in stock market

Stochastic Gradient Descent Training of Ensembles of DT-CNN Classifiers for Digit Recognition

A software toolbox for constructing ensembles of heterogenous linear and nonlinear models