SEARCH · PubMed Health

Results for “Multilayer Perceptrons”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 145 records · Page 8Linked to original sources

PubMed2000

Artificial neural networks in liquid chromatography: efficient and improved quantitative structure-retention relationship models.

The application of the principal neural network architecture, namely the multilayer perceptron (MLP), has been developed for obtaining sufficient quantitative structure-retention relationships (QSRR) with high accuracy. The present study is an extension to the excellent study of Cserhati et al. [LC-GC Int., 11 (1998) 240] for the retention behavior of solutes based on their structure. To this end, a dataset of 25 substances as solutes to two different stationary phases (polyethylene-silica and polyethylene-alumina) were analyzed to their structural descriptors and related to their retention behavior as expressed by the logarithms of their capacity factors (log k'). The results were compared to those of Cserhati et al. who studied the same problem using as many as ten different equations based on multiple regression analysis. In the present study a series of new and improved algorithms other than the 'old-fashioned' and problematic steepest descent were examined for training the MLP networks. The proposed methods led to substantial gain in both the prediction ability and the computation speed of the resulting models. For the development and evaluation of the artificial neural network (ANN) systems the same (eight) descriptors proposed by Cserhati were used also in this study. Furthermore, the results were compared to those produced from classical linear multivariate regression such as partial least squares regression (PLS). Some of the proposed ANN models diminished the number of outliers, during their implementation to unseen data (solutes), to zero.

Algorithms↗

PubMed2003

An artificial neural network to predict the outcome of repeat prostate biopsies.

OBJECTIVES: To develop an advanced artificial neural network (ANN) to predict the presence of prostate cancer (PCa) and to predict the outcome of repeat prostate biopsies. The predictive accuracy was compared with the accuracy obtained using standard cutoffs for the free/total (f/t) prostate-specific antigen (PSA) ratio, PSA density (PSAD), PSA density of the transition zone (PSA-TZ), and the total and transition zone volumes. Clinical and biochemical diagnostic tests have been shown to improve PCa detection. When these tests are combined using an ANN, significant increases in specificity at high sensitivity are observed. METHODS: The Vienna-based multicenter European referral database for early PCa detection of 820 men with a PSA level between 4 and 10 ng/mL was used. The presence of PCa was determined using transrectal ultrasound-guided octant needle repeat biopsy. Variables in the database consisted of age, PSA, f/t PSA ratio, digital rectal examination findings, PSA velocity, and the transrectal ultrasound-guided variables of prostate volume, transition zone volume, PSAD, and PSA-TZ. The ANN used in the analysis was an advanced multilayer perceptron selected for accuracy by a genetic algorithm. RESULTS: The repeat biopsy PCa detection rate was 10% (n = 83). At 95% sensitivity, the specificity for ANN was 68% compared with 54%, 33.5%, 21.4%, 14.7%, and 8.3% for multivariate logistic regression analysis, f/t PSA ratio, PSA-TZ, PSAD, and total PSA, respectively. The ANN reduced unnecessary repeat biopsies by 68% in this study. The area under the curve was 83% for the ANN versus 79%, 74.5%, 69.1%, 61.8%, and 60.5% for multivariate analysis, f/t PSA ratio, PSA-TZ, PSAD, and total PSA, respectively. CONCLUSIONS: The current ANN found a strong pattern predictive of PCa in patients with a negative initial biopsy. By combining the individual clinical and biochemical markers into the ANN, 68% specificity at 95% sensitivity was achieved. The ANN allows more accurate and individual counseling of patients with a negative initial biopsy.

Aged↗

PubMed2002

New approach for modelling ultrasound blood backscatter signal.

The ultrasound (US) scattered signal from blood has been treated as a random signal by many investigators. However, the degree of randomness of a medium is a relative term that can change considerably with the resolution of the sensor. In this study, the backscattered signal from blood has been looked at as a chaotic signal. By this treatment, according to Taken's theorem, a single variable (e.g., amplitude of the blood-backscattered signal) can be used to reconstruct the nonlinear dynamics of the blood-scattered signal. Multilayer perceptron neural network architecture, with error back-propagation, has been formulated and used as a basis for building and testing the chaotic model of the backscattered signal. This chaotic model is used successfully as a short-term predictor of the backscattered signal from blood-mimicking fluid (BMF) flowing in a vascular flow phantom under pulsatile flow. This modelling approach can be useful, for example, in detecting blood-borne emboli.

Blood↗

PubMed2002

Diagnosis of breast tumors with sonographic texture analysis using wavelet transform and neural networks.

To increase the ability of ultrasonographic technology for the differential diagnosis of solid breast tumors, we describe a novel computer-aided diagnosis (CADx) system using neural networks for classification of breast tumors. Tumor regions and surrounding tissues are segmented from the physician-located region-of-interest (ROI) images by applying our proposed segmentation algorithm. Cooperating with the segmentation algorithm, three feasible features, including variance contrast, autocorrelation contrast and distribution distortion of wavelet coefficients, were extracted from the ROI images for further classification. A multilayered perceptron (MLP) neural network trained using error back-propagation algorithm with momentum was then used for the differential diagnosis of breast tumors on sonograms. In the experiment, 242 cases (including benign breast tumors from 161 patients and carcinomas from 82 patients) were sampled with k-fold cross-validation (k = 10) to evaluate the performance. The receiver operating characteristic (ROC) area index for the proposed CADx system is 0.9396 +/- 0.0183, the sensitivity is 98.77%, the specificity is 81.37%, the positive predictive value is 72.73% and the negative predictive value is 99.24%. Experimental results showed that our diagnosis model performed very well for breast tumor diagnosis.

Breast Diseases↗

PubMed2003

Artificial neural networks for species identification by taxonomists.

This paper is a study of the value of applying artificial neural networks (ANNs), specifically a multilayer perceptron (MLP), to identification of higher plants using morphological characters collected by conventional means. A practical methodology is thus demonstrated to enable botanical or zoological taxonomists to use ANNs as advisory tools for identification purposes. A comparison is made between the ability of the neural network and that of traditional methods for plant identification by means of a case study in the flowering plant genus Lithops N.E. Brown (Aizoaceae). In particular, a comparison is made with taxonomic keys generated by means of the DELTA system. The ANN is found to perform better than the DELTA key generator, for conditions where the available data is limited, and species relatively difficult to distinguish.

Aizoaceae↗

PubMed2001

Quantitative structure-binding relationships (QSBR) and artificial neural networks: improved predictions in drug:cyclodextrin inclusion complexes.

The application of the principal neural network architecture, namely the multilayer perceptron (MLP), have been developed for obtaining sufficient quantitative structure-binding relationships (QSBR) with high accuracy. To this end a dataset of 17 barbiturates as guests complexing to alpha- and beta-cyclodextrins was examined and the results compared to that of Lopata et al (J. Pharm. Sci., 74, (1995)) who studied the same problem using multiple regression analysis. A series of new and improved algorithms other than the "old fashion" and problematic steepest descent were examined for training the MLP networks. The proposed methods led to substantial gain in both the prediction ability and the computation speed of the resulting models. A specific ANN architecture (4-4-1) trained with the Levenberg-Marquardt algorithm diminished the number of outliers, during its implementation to unseen data (barbiturates), to zero.

Algorithms↗

PubMed2001

Three learning phases for radial-basis-function networks.

In this paper, learning algorithms for radial basis function (RBF) networks are discussed. Whereas multilayer perceptrons (MLP) are typically trained with backpropagation algorithms, starting the training procedure with a random initialization of the MLP's parameters, an RBF network may be trained in many different ways. We categorize these RBF training methods into one-, two-, and three-phase learning schemes. Two-phase RBF learning is a very common learning scheme. The two layers of an RBF network are learnt separately; first the RBF layer is trained, including the adaptation of centers and scaling parameters, and then the weights of the output layer are adapted. RBF centers may be trained by clustering, vector quantization and classification tree algorithms, and the output layer by supervised learning (through gradient descent or pseudo inverse solution). Results from numerical experiments of RBF classifiers trained by two-phase learning are presented in three completely different pattern recognition applications: (a) the classification of 3D visual objects; (b) the recognition hand-written digits (2D objects); and (c) the categorization of high-resolution electrocardiograms given as a time series (ID objects) and as a set of features extracted from these time series. In these applications, it can be observed that the performance of RBF classifiers trained with two-phase learning can be improved through a third backpropagation-like training phase of the RBF network, adapting the whole set of parameters (RBF centers, scaling parameters, and output layer weights) simultaneously. This, we call three-phase learning in RBF networks. A practical advantage of two- and three-phase learning in RBF networks is the possibility to use unlabeled training data for the first training phase. Support vector (SV) learning in RBF networks is a different learning approach. SV learning can be considered, in this context of learning, as a special type of one-phase learning, where only the output layer weights of the RBF network are calculated, and the RBF centers are restricted to be a subset of the training data. Numerical experiments with several classifier schemes including k-nearest-neighbor, learning vector quantization and RBF classifiers trained through two-phase, three-phase and support vector learning are given. The performance of the RBF classifiers trained through SV learning and three-phase learning are superior to the results of two-phase learning, but SV learning often leads to complex network structures, since the number of support vectors is not a small fraction of the total number of data points.

Algorithms↗

PubMedDate not supplied

A spiking neural network architecture for nonlinear function approximation.

Multilayer perceptrons have received much attention in recent years due to their universal approximation capabilities. Normally, such models use real valued continuous signals, although they are loosely based on biological neuronal networks that encode signals using spike trains. Spiking neural networks are of interest both from a biological point of view and in terms of a method of robust signaling in particularly noisy or difficult environments. It is important to consider networks based on spike trains. A basic question that needs to be considered however, is what type of architecture can be used to provide universal function approximation capabilities in spiking networks? In this paper, we propose a spiking neural network architecture using both integrate-and-fire units as well as delays, that is capable of approximating a real valued function mapping to within a specified degree of accuracy.

Action Potentials↗

PubMed2001

A closed-form neural network for discriminatory feature extraction from high-dimensional data.

We consider a new neural network for data discrimination in pattern recognition applications. We refer to this as a maximum discriminating feature (MDF) neural network. Its weights are obtained in closed-form, thereby overcoming problems associated with other nonlinear neural networks. It uses neuron activation functions that are dynamically chosen based on the application. It is theoretically shown to provide nonlinear transforms of the input data that are more general than those provided by other nonlinear multilayer perceptron neural network and support-vector machine techniques for cases involving high-dimensional (image) inputs where training data are limited and the classes are not linearly separable. We experimentally verify this on synthetic examples.

Algorithms↗

PubMed2002

A methodology to explain neural network classification.

Neural networks are still frustrating tools in the data mining arsenal. They exhibit excellent modelling performance, but do not give a clue about the structure of their models. We propose a methodology to explain the classification obtained by a multilayer perceptron. We introduce the concept of 'causal importance' and define a saliency measurement allowing the selection of relevant variables. Once the model is trained with the relevant variables only, we define a clustering of the data built from the hidden layer representation. Combining the saliency and the causal importance on a cluster by cluster basis allows an interpretation of the neural network classifier to be built. We illustrate the performances of this methodology on three benchmark datasets.

Classification↗

PubMed2002

A general framework for neural network models on censored survival data.

Flexible parametric techniques for regression analysis, such as those based on feed forward artificial neural networks (FFANNs), can be useful for the statistical analysis of censored time data. These techniques are of particular interest for the study of the outcome dependence from several variables measured on a continuous scale, since they allow for the detection of complex non-linear and non-additive effects. Few efforts have been made until now to account for censored times in FFANNs. In the attempt to fill this gap, specific error functions and data representation will be introduced for multilayer perceptron and radial basis function extensions of generalized linear models for survival data.

Humans↗

PubMedDate not supplied

Measuring solar magnetic fields with artificial neural networks.

The quantification of the solar magnetic field is a crucial step in modern solar physics to understand the dynamics, activity and variability of our star. Presently, a reliable inference of these fields is only possible by means of a computer-intensive process that has so far limited scientists to the analysis of observations from small regions of the solar disk, and/or very crude spatial and temporal resolution. This work presents a different approach to the problem, in which a multilayer perceptron, trained with known synthetic profiles, is able to recognize the profiles and return the magnetic field used to synthesize them. The network is then confronted with real observations of a sunspot which had been previously inverted using traditional inversion techniques. A quantitative comparison between these two procedures shows the reliability of the network when applied to points having magnetic filling factors larger than approximately 70%. The dramatic decrease in the required computing time presents an opportunity for the routine analysis of large-scale, high-resolution solar observations.

Electromagnetic Fields↗

PubMedDate not supplied

On structure-exploiting trust-region regularized nonlinear least squares algorithms for neural-network learning.

This paper briefly introduces our numerical linear algebra approaches for solving structured nonlinear least squares problems arising from 'multiple-output' neural-network (NN) models. Our algorithms feature trust-region regularization, and exploit sparsity of either the 'block-angular' residual Jacobian matrix or the 'block-arrow' Gauss-Newton Hessian (or Fisher information matrix in statistical sense) depending on problem scale so as to render a large class of NN-learning algorithms 'efficient' in both memory and operation costs. Using a relatively large real-world nonlinear regression application, we shall explain algorithmic strengths and weaknesses, analyzing simulation results obtained by both direct and iterative trust-region algorithms with two distinct NN models: 'multilayer perceptrons' (MLP) and 'complementary mixtures of MLP-experts' (or neuro-fuzzy modular networks).

Algorithms↗

PubMedDate not supplied

Divide-and-conquer approach for brain machine interfaces: nonlinear mixture of competitive linear models.

This paper proposes a divide-and-conquer strategy for designing brain machine interfaces. A nonlinear combination of competitively trained local linear models (experts) is used to identify the mapping from neuronal activity in cortical areas associated with arm movement to the hand position of a primate. The proposed architecture and the training algorithm are described in detail and numerical performance comparisons with alternative linear and nonlinear modeling approaches, including time-delay neural networks and recursive multilayer perceptrons, are presented. This new strategy allows training the local linear models using normalized LMS and using a relatively smaller nonlinear network to efficiently combine the predictions of the linear experts. This leads to savings in computational requirements, while the performance is still similar to a large fully nonlinear network.

Artificial Intelligence↗

PubMed1997

Training Dynamics and Neural Network Performance.

We use an analysis of a simple model of recurrent network dynamics to gain qualitative insights into the training dynamics of feedforward multilayer perceptrons (MLPs) used for classification. These insights suggest changes to the training methods used for MLPs that improve network performance significantly. In previous work, the probabilistic neural network (PNN) was shown to provide better zero-reject error performance on character and fingerprint classification problems than radial basis function and MLP-based neural network methods. We will show that performance equal to or better than PNN can be achieved with a single three-layer MLP by making fundamental changes in the network optimization strategy. These changes are: 1) use of neuron activation functions, which reduce the probability of singular Jacobians; 2) use of successive regularization to constrain the volume of the minimized weight space; 3) use of Boltzmann pruning to constrain the dimension of the weight space; 4) use of Prior class probabilities to normalize all error calculations, so that statistically significant samples of rare but important classes can be included without distorting the error surface. All four of these changes are made in the inner loop of a conjugate gradient optimization iteration and are intended to simplify the training dynamics of the optimization. On handprinted digits and fingerprint classification problems these modifications improve error-reject performance by factors between 2 and 4, and reduce network size by 40-60%. Copyright 1997 Elsevier Science Ltd.

Journal Article↗

PubMed1997

Distributed Learning, Recognition, and Prediction by ART and ARTMAP Neural Networks.

A class of adaptive resonance theory (ART) models for learning, recognition, and prediction with arbitrarily distributed code representations is introduced. Distributed ART neural networks combine the stable fast learning capabilities of winner-take-all ART systems with the noise tolerance and code compression capabilities of multilayer perceptrons. With a winner-take-all code, the unsupervised model dART reduces to fuzzy ART and the supervised model dARTMAP reduces to fuzzy ARTMAP. With a distributed code, these networks automatically apportion learned changes according to the degree of activation of each coding node, which permits fast as well as slow learning without catastrophic forgetting. Distributed ART models replace the traditional neural network path weight with a dynamic weight equal to the rectified difference between coding node activation and an adaptive threshold. Thresholds increase monotonically during learning according to a principle of atrophy due to disuse. However, monotonic change at the synaptic level manifests itself as bidirectional change at the dynamic level, where the result of adaptation resembles long-term potentiation (LTP) for single-pulse or low frequency test inputs but can resemble long-term depression (LTD) for higher frequency test inputs. This paradoxical behavior is traced to dual computational properties of phasic and tonic coding signal components. A parallel distributed match-reset-search process also helps stabilize memory. Without the match-reset-search system, dART becomes a type of distributed competitive learning network.

Journal Article↗

PubMed1997

Common Optimization of Adaptive Preprocessing Units and a Neural Network during the Learning Period. Application in EEG Pattern Recognition.

In this study, a proposition of simultaneous training of the neural network (multilayer perceptron) and adaptive preprocessing unit is presented. This cooperation enables the network to affect the preprocessing and as a consequence to vary the locations of pattern vectors in a feature space. Thus, during the learning process the network tries to find a good separation of classes of patterns, which results in convergence of the whole learning process. The strategy was developed in order to make efficient EEG monitoring in neonates possible. A comparison of the method presented herein with the known learning strategies for neural networks shows the need for using it as an alternative learning process. The convergence of the whole system is also discussed. Copyright 1997 Elsevier Science Ltd.

Journal Article↗

PubMed1998

A second-order learning algorithm for multilayer networks based on block Hessian matrix.

This article proposes a new second-order learning algorithm for training the multilayer perceptron (MLP) networks. The proposed algorithm is a revised Newton's method. A forward-backward propagation scheme is first proposed for network computation of the Hessian matrix, H, of the output error function of the MLP. A block Hessian matrix, H(b), is then defined to approximate and simplify H. Several lemmas and theorems are proved to uncover the important properties of H and H(b), and verify the good approximation of H(b) to H; H(b) preserves the major properties of H. The theoretic analysis leads to the development of an efficient way for computing the inverse of H(b) recursively. In the proposed second-order learning algorithm, the least squares estimation technique is adopted to further lessen the local minimum problems. The proposed algorithm overcomes not only the drawbacks of the standard backpropagation algorithm (i.e. slow asymptotic convergence rate, bad controllability of convergence accuracy, local minimum problems, and high sensitivity to learning constant), but also the shortcomings of normal Newton's method used on the MLP, such as the lack of network implementation of H, ill representability of the diagonal terms of H, the heavy computation load of the inverse of H, and the requirement of a good initial estimate of the solution (weights). Several example problems are used to demonstrate the efficiency of the proposed learning algorithm. Extensive performance (convergence rate and accuracy) comparisons of the proposed algorithm with other learning schemes (including the standard backpropagation algorithm) are also made.

Journal Article↗