Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Neural network”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

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

At least 235 records · Page 13Linked to original sources
PubMed1999

Modular neural networks: a survey.

Modular Neural Networks (MNNs) is a rapidly growing field in artificial Neural Networks (NNs) research. This paper surveys the different motivations for creating MNNs: biological, psychological, hardware, and computational. Then, the general stages of MNN design are outlined and surveyed as well, viz., task decomposition techniques, learning schemes and multi-module decision-making strategies. Advantages and disadvantages of the surveyed methods are pointed out, and an assessment with respect to practical potential is provided. Finally, some general recommendations for future designs are presented.

Biology↗
PubMedDate not supplied

Bayesian neural networks for aroma classification.

Bayesian Neural Networks (BNNs) are investigated to test their potential to distinguish between different aroma impressions. Special attention is thereby drawn on mixed aroma impressions, resulting from the flavor description of a single compound with more than one aroma quality. The structures of 133 pyrazine-derived aroma compounds as well as their aroma descriptions are selected for comparison. The information fed into the neural networks is based on molecular descriptors calculated from the geometrically optimized chemical structures. While in the case of the Probabilistic Neural Network (PNN) the networks' output consists of a categorical variable, the output for the General Regression Neural Network (GRNN) is defined in a numerical way. The best models attain comparable performance with a correct prediction of 90.8% of the cases for PNN and 89.9% for GRNN, respectively. Comparison of the BNN results to those obtained by Multiple Linear Regression (MLR) points out that the nonlinear methods work significantly better on the studied problem and that BNNs can be applied to multiple-category problems in structure-flavor relationships with good accuracy.

Bayes Theorem↗
PubMed1992

Clustering proteins into families using artificial neural networks.

An artificial neural network was used to cluster proteins into families. The network, composed of 7 x 7 neurons, was trained with the Kohonen unsupervised learning algorithm using, as inputs, matrix patterns derived from the bipeptide composition of 447 proteins, belonging to 13 different families. As a result of the training, and without any a priori indication of the number or composition of the expected families, the network self-organized the activation of its neurons into topologically ordered maps in which almost all the proteins (96.7%) were correctly clustered into the corresponding families. In a second computational experiment, a similar network was trained with one family of the previous learning set (76 cytochrome c sequences). The new neural map clustered these proteins into 25 different neurons (five in the first experiment), wherein phylogenetically related sequences were positioned close to each other. This result shows that the network can adapt the clustering resolution to the complexity of the learning set, a useful feature when working with an unknown number of clusters. Although the learning stage is time consuming, once the topological map is obtained, the classification of new proteins is very fast. Altogether, our results suggest that this novel approach may be a useful tool to organize the search for homologies in large macromolecular databases.

Algorithms↗
PubMedDate not supplied

Potential usefulness of an artificial neural network for assessing ventricular size.

An artificial neural network approach was applied to assess ventricular size from computed tomograms. Three layer, feed-forward neural networks with a back propagation algorithm were designed to distinguish between three degrees of enlargement of the ventricles on the basis of patient's age and six items of computed tomographic information. Data for training and testing the neural network were created with computed tomograms of the brains selected at random from daily examinations. Four radiologists decided by mutual consent subjectively based on their experience whether the ventricles were within normal limits, slightly enlarged, or enlarged for the patient's age. The data for training was obtained from 38 patients. The data for testing was obtained from 47 other patients. The performance of the neural network trained using the data for training was evaluated by the rate of correct answers to the data for testing. The valid solution ratio to response of the test data obtained from the trained neural networks was more than 90% for all conditions in this study. The solutions were completely valid in the neural networks with two or three units at the hidden layer with 2,200 learning iterations, and with two units at the hidden layer with 11,000 learning iterations. The squared error decreased remarkably in the range from 0 to 500 learning iterations, and was close to constant over two thousand learning iterations. The neural network with a hidden layer having two or three units showed high decision performance. The preliminary results strongly suggest that the neural network approach has potential utility in computer-aided estimation of enlargement of the ventricles.

Cerebral Ventricles↗
PubMed2005

Autoassociative memory design using interconnected generalized brain-state-in-a-box neural networks.

A class of interconnected neural networks composed of generalized Brain-State-in-a-Box (gBSB) neural subnetworks is considered. Interconnected gBSB neural network architectures are proposed along with their stability conditions. The design of the interconnected neural networks is reduced to the problem of solving linear matrix inequalities (LMIs) to determine the interconnection parameters. A method for solving LMIs is devised generating the solutions that, in general, are further away from zero than the corresponding solutions obtained using MATLAB's LMI toolbox, thus resulting in stronger interconnections between the subnetworks. The proposed architectures are then used to construct neural associative memories. Simulations are performed to illustrate the results obtained.

Algorithms↗
PubMed1999

Artificial neural networks as approximators of stochastic processes.

Artificial Neural Networks (ANNs) must be able to learn by experience from environment. This property can be considered as being closely related to the approximating capabilities of the networks. Unfortunately at present only the ability of ANNs in approximating deterministic input-output mappings has been exploited. In this article it has been shown that some classes of neural networks, named Stochastic Neural Networks, which are capable of using approximating stochastic processes are defined. As stochastic processes may also be viewed as random functions, they include deterministic (non-random) functions as a particular case. Thus the class of Stochastic Neural Networks can be considered as a generalisation of the usually defined neural networks. From an application point of view such a class of networks is more adherent to real world in which neural networks must work in an environment which is essentially stochastic. The theory presented in the article has been carried out starting from the so-called "canonical representation" for non-stationary stochastic processes. Finally, an application example showing in detail the validity of the proposed approach has been reported.

Journal Article↗
PubMed2003

Estimation of pulmonary artery occlusion pressure by an artificial neural network.

OBJECTIVE: We hypothesized that an artificial neural network, interconnected computer elements capable of adaptation and learning, could accurately estimate pulmonary artery occlusion pressure from the pulsatile pulmonary artery waveform. SETTING: University medical center. SUBJECTS: Nineteen closed-chest dogs. INTERVENTIONS: Pulmonary artery waveforms were digitally sampled before conventional measurements of pulmonary artery occlusion pressure under control conditions, during infusions of serotonin or histamine, or during volume loading. Individual beats were parsed or separated out. Pulmonary artery pressure, its first time derivative, and the beat duration were used as neural inputs. The neural network was trained by using 80% of all samples and tested on the remaining 20%. For comparison, the regression between pulmonary artery diastolic pressure and pulmonary artery occlusion pressure was developed and tested using the same data sets. As a final test of generalizability, the neural network was trained on data obtained from 18 dogs and tested on data from the remaining dog in a round-robin fashion. MEASUREMENTS AND MAIN RESULTS: The correlation coefficient between the pulmonary artery diastolic pressure estimate of pulmonary artery occlusion pressure and measured pulmonary artery occlusion pressure was.75, whereas that for the neural network estimate of pulmonary artery occlusion pressure was.97 (p <.01 for difference between pulmonary artery diastolic pressure and pulmonary artery occlusion pressure estimates). The pulmonary artery diastolic pressure estimate of pulmonary artery occlusion pressure showed a bias of 0.097 mm Hg (limits of agreement -7.57 to 7.767 mm Hg), whereas the neural network estimate of pulmonary artery occlusion pressure showed a bias of -0.002 mm Hg (-2.592 to 2.588 mm Hg). There was no significant change in the bias of the neural network estimate over the range of values tested. In contrast, the bias for the pulmonary artery diastolic pressure estimate significantly increased with the increasing magnitude of the pulmonary artery occlusion pressure. During round-robin testing, the neural network estimate of pulmonary artery occlusion pressure showed suboptimal performance (correlation coefficient between estimated and measured pulmonary artery occlusion pressure.59). CONCLUSIONS: A neural network can accurately estimate pulmonary artery occlusion pressure over a wide range of pulmonary artery occlusion pressure under conditions that alter pulmonary hemodynamics. We speculate that artificial neural networks could provide accurate, real-time estimates of pulmonary artery occlusion pressure in critically ill patients.

Animals↗
PubMed2002

Validation using sensitivity and target transform factor analyses of neural network models for classifying bacteria from mass spectra.

Temperature constrained cascade correlation networks (TCCCNs) are computational neural networks that configure their own architecture, train rapidly, and give reproducible prediction results. TCCCN classification models were built using the Latin-partition method for five classes of pathogenic bacteria. Neural networks are problematic in that the relationships among the inputs (i.e., mass spectra) and the outputs (i.e., the bacterial identities) are not apparent. In this study, neural network models were constructed that successfully classified the targeted bacteria and the classification model was validated using sensitivity and target transformation factor analysis (TTFA). Without validation of the classification model, it is impossible to ascertain whether the bacteria are classified by peaks in the mass spectrum that have no causal relationships with the bacteria, but instead randomly correlate with the bacterial classes. Multiple single output network models did not offer any benefits when compared to single network models that had multiple outputs. A multiple output TCCCN model achieved classification accuracies of 96 +/- 2% and exhibited improved performance over multiple single output TCCCN models. Chemical ionization mass spectra were obtained from in situ thermal hydrolysis methylation of freeze-dried bacteria. Mass spectral peaks that pertain to the neural network classification model of the pathogenic bacterial classes were obtained by sensitivity analysis. A significant number of mass spectral peaks that had high sensitivity corresponded to known biomarkers, which is the first time that the significant peaks used by a neural network model to classify mass spectra have been divulged. Furthermore, TTFA furnishes a useful visual target as to which peaks in the mass spectrum correlate with the bacterial identities.

Bacteria↗
PubMed2003

On learning to estimate the block directional image of a fingerprint using a hierarchical neural network.

This paper presents a hierarchical neural network architecture for computing fingerprints block directional images. Two separately trained neural networks are connected in series. First, the fingerprint image is divided into 16x16 blocks, each block is submitted to the first network which is a back propagation neural network. It has four counters in its output layer one for each direction to count the main directional codes in each fingerprint block. The output of this network is considered the feature vector for the fingerprint block, which is then submitted to the second network. The second network is a self-organized feature maps neural network uses an unsupervised learning strategy to group the fingerprint blocks into distinct directional classes. In this scheme, there is more than one sub-class for each directional class, an agglomerative hierarchical cluster algorithm for merging two clusters is used to merge two classes if their corresponding distances are below a specified threshold. Results obtained with a real world data set indicate the effectiveness of the proposed architecture.

Algorithms↗
PubMed2003

Intelligent optimal control with dynamic neural networks.

The application of neural networks technology to dynamic system control has been constrained by the non-dynamic nature of popular network architectures. Many of difficulties are-large network sizes (i.e. curse of dimensionality), long training times, etc. These problems can be overcome with dynamic neural networks (DNN). In this study, intelligent optimal control problem is considered as a nonlinear optimization with dynamic equality constraints, and DNN as a control trajectory priming system. The resulting algorithm operates as an auto-trainer for DNN (a self-learning structure) and generates optimal feed-forward control trajectories in a significantly smaller number of iterations. In this way, optimal control trajectories are encapsulated and generalized by DNN. The time varying optimal feedback gains are also generated along the trajectory as byproducts. Speeding up trajectory calculations opens up avenues for real-time intelligent optimal control with virtual global feedback. We used direct-descent-curvature algorithm with some modifications (we called modified-descend-controller-MDC algorithm) for the optimal control computations. The algorithm has generated numerically very robust solutions with respect to conjugate points. The adjoint theory has been used in the training of DNN which is considered as a quasi-linear dynamic system. The updating of weights (identification of parameters) are based on Broyden-Fletcher-Goldfarb-Shanno BFGS method. Simulation results are given for an intelligent optimal control system controlling a difficult nonlinear second-order system using fully connected three-neuron DNN.

Artificial Intelligence↗
PubMed2005

An improved genetic algorithm based fuzzy-tuned neural network.

This paper presents a fuzzy-tuned neural network, which is trained by an improved genetic algorithm (GA). The fuzzy-tuned neural network consists of a neural-fuzzy network and a modified neural network. In the modified neural network, a neuron model with two activation functions is used so that the degree of freedom of the network function can be increased. The neural-fuzzy network governs some of the parameters of the neuron model. It will be shown that the performance of the proposed fuzzy-tuned neural network is better than that of the traditional neural network with a similar number of parameters. An improved GA is proposed to train the parameters of the proposed network. Sets of improved genetic operations are presented. The performance of the improved GA will be shown to be better than that of the traditional GA. Some application examples are given to illustrate the merits of the proposed neural network and the improved GA.

Algorithms↗
PubMed1996

Population networks: a large-scale framework for modelling cortical neural networks.

Artificial neural networks are usually built on rather few elements such as activation functions, learning rules, and the network topology. When modelling the more complex properties of realistic networks, however, a number of higher-level structural principles become important. In this paper we present a theoretical framework for modelling cortical networks at a high level of abstraction. Based on the notion of a population of neurons, this framework can accommodate the common features of cortical architecture, such as lamination, multiple areas and topographic maps, input segregation, and local variations of the frequency of different cell types (e.g., cytochrome oxidase blobs). The framework is meant primarily for the simulation of activation dynamics; it can also be used to model the neural environment of single cells in a multiscale approach.

Animals↗
PubMed2000

Predicting pediatric length of stay and acuity of care in the first ten minutes with artificial neural networks.

OBJECTIVE: To evaluate the efficacy of artificial neural networks in categorizing pediatric trauma patients into four distinct acuity of care groups and in determining the length of stay (LOS) within specific areas of the hospital. DESIGN: Using historical information from >8,000 pediatric trauma patient records, train and evaluate artificial neural networks to predict the injury severity and LOS for each patient in pediatric intensive care units (PICUs), step-down units, and floor units. Each artificial neural network is evaluated for categorization accuracy and mean absolute error difference on the predicted LOS. SUBJECTS: A total of 10,353 patient records from the National Pediatric Trauma registry, representing all pediatric trauma patients treated at affiliated hospitals from April 1994 through December 1996. Records with incomplete information were eliminated from the study, leaving 8,081 usable patient records. MEASUREMENTS: A total of 14 variables are selected from the 81 values present in the National Pediatric Trauma Registry as independent variables for the artificial neural networks. Each neural network produces nine output values: five for categorizing the patient's injury severity, three for the LOS in the PICU, step-down unit, and floor units, and one for the patient's total LOS. RESULTS: A fuzzy ARTMAP neural network accurately categorizes 88% of mortality patients and 58.3% of critical PICU patients. A backpropagation neural network succeeded in predicting the total LOS to within 1 day for 51.4% and the ICU LOS to within 1 day for 70.4% of all evaluated patients. CONCLUSION: Information available in the first 10 mins of a patient's presentation at the emergency room can be used by an artificial neural network to predict injury severity and LOS. Artificial neural networks enable more effective resource planning and patient management.

Journal Article↗
PubMed1991

Prediction of ATP/GTP-binding motif: a comparison of a perceptron type neural network and a consensus sequence method [corrected].

Neural networks have been applied to a number of protein structure problems. In some applications their success has not been substantiated by a comparison with the performance of a suitable alternative statistical method on the same data. In this paper, a two-layer feed-forward neural network has been trained to recognize ATP/GTP-binding [corrected] local sequence motifs. The neural network correctly classified 78% of the 349 sequences used. This was much better than a simple motif-searching program. A more sophisticated statistical method was developed, however, which performed marginally better (80% correct classification) than the neural network. The neural network and the statistical method performed similarly on sequences of varying degrees of homology. These results do not imply that neural networks, especially those with hidden layers, are not useful tools, but they do suggest that two-layer networks in particular should be carefully tested against other statistical methods.

Adenosine Triphosphate↗
PubMedDate not supplied

Structure-based predictions of 1H NMR chemical shifts using feed-forward neural networks.

Feed-forward neural networks were trained for the general prediction of 1H NMR chemical shifts of CH(n) protons in organic compounds in CDCl3. The training set consisted of 744 1H NMR chemical shifts from 120 molecular structures. The method was optimized in terms of selected proton descriptors (selection of variables), the number of hidden neurons, and integration of different networks in ensembles. Predictions were obtained for an independent test set of 952 cases with a mean average error of 0.29 ppm (0.20 ppm for 90% of the cases). The results were significantly better than those obtained with counterpropagation neural networks.

Hydrogen↗
PubMed1996

Artificial neural network prediction of ascites in broilers.

An artificial neural network was trained to predict the presence or absence of ascites in broiler chickens. The neural network was a three-layer back-propagation neural network with an input layer of 15 neurons (defining 15 physiological variables), a hidden layer of 16 neurons, and an output layer of 2 neurons (the presence or absence of ascites). Male by-products of a breeder pullet line were brooded at 32 and 30 C during Weeks 1 and 2, respectively. The training set for the neural network consisted of data from birds subjected to cool temperatures (18 C) to induce ascites. After training, the predictive ability of the neural network was verified with two new data sets. The second data set was from birds subjected to cool temperatures (18 C). The third data set was from birds subjected to clamping of the pulmonary artery to simulate the physiological processes involved in ascites (the temperature was 24 C). A comparison was made between laboratory diagnostic results and the neural network predicted ascites incidence. The neural network accurately identified the presence or absence of ascites in the first (training) set. Two false positives and one false positive were identified in the second and third verification sets, respectively. The birds identified as false positives were determined to be in the developmental stages of ascites before the occurrence of fluid accumulation. Artificial neural networks were found to effectively identify broilers with and without ascites.

Animals↗
PubMed1991

Prediction of protein secondary structure by an enhanced neural network.

Computational model of neural network is used for prediction of secondary structure of globular proteins of known sequence. In contrast to earlier works some information about expected tertiary interactions were built in into the neural network. As a result the prediction accuracy was improved by 3% to 5%. Possible applications of this new approach are briefly discussed.

Amino Acid Sequence↗
PubMed2005

A primal-dual neural network for online resolving constrained kinematic redundancy in robot motion control.

This paper proposes a primal-dual neural network with a one-layer structure for online resolution of constrained kinematic redundancy in robot motion control. Unlike the Lagrangian network, the proposed neural network can handle physical constraints, such as joint limits and joint velocity limits. Compared with the existing primal-dual neural network, the proposed neural network has a low complexity for implementation. Compared with the existing dual neural network, the proposed neural network has no computation of matrix inversion. More importantly, the proposed neural network is theoretically proved to have not only a finite time convergence, but also an exponential convergence rate without any additional assumption. Simulation results show that the proposed neural network has a faster convergence rate than the dual neural network in effectively tracking for the motion control of kinematically redundant manipulators.

Algorithms↗