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 109 records · Page 6Linked to original sources
PubMed2005

Modeling the electrophoretic mobility of beta-blockers in capillary electrophoresis using artificial neural networks.

Artificial neural networks were used for modeling the mobility of five beta-blockers (i.e., labetalol atenolol, practolol, timolol and propranolol) in running buffer with ternary solvent background electrolyte systems containing 80 mM acetate buffer dissolved in water, methanol, ethanol and their ternary mixtures. The volume fractions of two solvents (f(2), f(3)) and cologarithm of electrophoretic mobilities in pure solvents (i.e., -Lnmu(1), -Lnmu(2) and -Lnmu(3)) were used as inputs and cologarithm of the mobility in mixed solvents was the output of the networks. The number of neurons in hidden layer, learning rate, momentum and the number of epochs were optimized, in which two neurons in hidden layer, 0.2, 0.9 and 20000 were found the optimized values for learning rate, momentum and number of epochs, respectively. Mean percentage deviations (MPD) between calculated and experimental mobilities were computed as an accuracy criterion. To assess the correlative ability of the model, all data points in each set were used as training set and the mobilities were back-calculated by the trained networks, in which the overall MPD (OMPD)+/- standard deviation (SD) for correlative study was 3.1+/- 2.3. To evaluate the prediction capability of the proposed ANN model, the network was trained using 15 data points for each analyte and the remaining data points were predicted. The obtained OMPD (+/-SD) for this analysis was 3.6+/-3.0. To further investigate on the applicability of ANN, a generalized network was trained with 10 data points from each beta-blocker and then the network was employed to predict the mobilities of the analytes in ternary solvent electrolyte systems. The MPDs for predicted mobilities were 3.6%, 3.6%, 3.9%, 3.7% and 2.9% respectively for labetalol, atenolol, practolol, timolol and propranolol.

Adrenergic beta-Antagonists↗
PubMed2003

Quantitative structure-property relationship studies of migration index in microemulsion electrokinetic chromatography using artificial neural network.

Artificial neural networks (ANNs) were successfully developed for the modeling and prediction of migration indices of the 53 benzene derivatives and heterocyclic compounds in microemulsion electrokinetic chromatography. The selected descriptors that appear in multiple linear regression models are: 3D-MoRSE signal 25 unweighted, 3D-MoRSE signal 19 weighted by atomic Sanderson electronegativity, R maximal autocorrelation index lag 1 weighted by atomic mass (R1M+), R maximal autocorrelation index lag 2 weighted by polarizability (R2P+) and average atomic composition index. These descriptors were used as inputs for generated 5-4-1 networks. After training and optimization of the ANN parameters it was used to prediction of migration index of the test set compounds. The results obtained using ANNs were compared with the experimental values as well as with those obtained using regression models and showed the superiority of ANNs over regression models.

Chromatography, Micellar Electrokinetic Capillary↗
PubMed2000

Predicting the standard enthalpy (deltaH0f) and entropy (S0) of alkanes by artificial neural networks.

Artificial Neural Networks (ANNs) with Extended Delta-Bar-Delta (EDBD) back propagation learning algorithm have been developed to predict the standard enthalpy and entropy of 87 acyclic alkanes. Molecular weight, boiling point and density of the compounds were used as input parameters. The network's architecture and parameters were optimized to give maximum performances. The best network was a 3-6-2 ANN, and the optimum learning epoch was about 1320. The results show that the maximum relative errors of enthalpy and entropy are less than 3%. They reveal that the performances of ANNs for predicting the enthalpy and entropy of alkanes are satisfying.

Algorithms↗
PubMed2000

Curved trajectory prediction using a self-organizing neural network.

Existing neural network models are capable of tracking linear trajectories of moving visual objects. This paper describes an additional neural mechanism, disfacilitation, that enhances the ability of a visual system to track curved trajectories. The added mechanism combines information about an object's trajectory with information about changes in the object's trajectory, to improve the estimates for the object's next probable location. Computational simulations are presented that show how the neural mechanism can learn to track the speed of objects and how the network operates to predict the trajectories of accelerating and decelerating objects.

Acceleration↗
PubMed1992

A movement pattern generator model using artificial neural networks.

Artificial neural networks (ANN's) allow a new approach to biological modeling. The main applications of ANN's have been geared towards the modeling of the association and learning mechanisms of the brain; only a few researchers have explored them for motor control. The fact that ANN's are based on biological systems indicates their potential application for a biological act such as locomotion. Towards this goal, we have developed a "movement pattern generator," using an ANN for generating periodic movement trajectories. This model is based on the concept of "central pattern generators." Jordan's sequential network, which is capable of learning sequences of patterns, was modified and used to generate several bipedal trajectories (or gaits), coded in task space, at different frequencies. The network model successfully learned all of the trajectories presented to it. The model has many attractive properties such as limit cycle behavior, generalization of trajectories and frequencies, phase maintenance, and fault tolerance. The movement pattern generator model is potentially applicable for improved understanding of animal locomotion and for use in legged robots and rehabilitation medicine.

Animals↗
PubMed1993

Fast learning of biased patterns in neural networks.

Usual neural network gradient descent training algorithms require training times of the same order as the number of neurons N if the patterns are biased. In this paper, modified algorithms are presented which require training times equal to those in unbiased cases which are of order 1. Exact convergence proofs are given. Gain parameters which produce minimal learning times in large networks are computed by replica methods. It is demonstrated how these modified algorithms are applied in order to produce four types of solutions to the learning problem: 1. A solution with all internal fields equal to the desired output, 2. The Adaline (or pseudo-inverse) solution, 3. The perceptron of optimal stability without threshold and 4. The perceptron of optimal stability with threshold.

Algorithms↗
PubMed2002

Learning and predicting time series by neural networks.

Artificial neural networks which are trained on a time series are supposed to achieve two abilities: first, to predict the series many time steps ahead and second, to learn the rule which has produced the series. It is shown that prediction and learning are not necessarily related to each other. Chaotic sequences can be learned but not predicted while quasiperiodic sequences can be well predicted but not learned.

Algorithms↗
PubMedDate not supplied

Constructing a query-able radial basis function artificial neural network.

Artificial neural networks will be more widely accepted as standard engineering tools if their reasoning process can be made less opaque. This paper describes NetQuery, an explanation mechanism that extracts meaningful explanations from trained Radial Basis Function (RBF) networks. RBF networks are well suited for explanation generation because they contain a set of locally tuned units. Standard RBF networks are modified to identify dependencies between the inputs, to be sparsely connected, and to have an easily interpretable output layer. Given these modifications, the network architecture can be used to extract "Why?" and "Why not?" explanations from the network in terms of excitatory and inhibitory in-puts and their linear relationships, greatly simplified by a run-time pruning algorithm. These query results are validated by creating an expert system based on the explanations. NetQuery is also able to inform a user about a possible change in category for a given pattern by responding to a "How can I...?" query. This kind of query is extremely useful when analyzing the quality of a pattern set.

Algorithms↗
PubMed1996

Prediction of lymph node metastasis in gastric cancer patients with neural networks.

Artificial neural networks are a kind of pattern classifiers, with growing acceptance in medical and biological research. We applied a single layer perceptron to data of 4302 patients from National Cancer Centre in Tokyo and compared the results to the Maruyama diagnostic system (MDS) and classic statistical analysis with logistic regression. While logistic regression reached no sensitivity and a specificity of 1.00 in median, MDS had a sensitivity of 0.74 and a specificity of 0.75 an median. The perceptron reached a median sensitivity of 0.81 and a median specificity of 0.71.

Adult↗
PubMed1997

Classification of user expertise level by neural networks.

A neural network approach to low-level user modeling is described, in the context of text editing tasks using the Jove editor. Knowledge of a user's expertise is extracted automatically, based on their interaction with Jove over a two week period. A MLP classifier which uses rprop learning and incorporates output data fuzzification is developed to classify users into one of five expertise levels. Classification into the correct level is achieved in around 80% of the cases, with misclassification being restricted to adjacent classes. The neurofuzzy system is seen to outperform not only the binary classifier of Beale [1989], but also production rule and inductive expert systems developed especially for comparison purposes in this study.

Artificial Intelligence↗
PubMed1999

Prediction of the location and type of beta-turns in proteins using neural networks.

A neural network has been used to predict both the location and the type of beta-turns in a set of 300 nonhomologous protein domains. A substantial improvement in prediction accuracy compared with previous methods has been achieved by incorporating secondary structure information in the input data. The total percentage of residues correctly classified as beta-turn or not-beta-turn is around 75% with predicted secondary structure information. More significantly, the method gives a Matthews correlation coefficient (MCC) of around 0.35, compared with a typical MCC of around 0.20 using other beta-turn prediction methods. Our method also distinguishes the two most numerous and well-defined types of beta-turn, types I and II, with a significant level of accuracy (MCCs 0.22 and 0.26, respectively).

Algorithms↗
PubMed2003

Sensor calibration and compensation using artificial neural network.

Artificial neural network (ANN) based inverse modeling technique is used for sensor response linearization. The choice of the order of the model and the number of the calibration points are important design parameters in this technique. An intensive study of the effect of the order of the model and number of calibration points on the lowest asymptotic root-mean-square (RMS) error has been reported in this paper. Starting from the initial value of the nonlinearity in the characteristics of a sensor and required RMS error, it is possible to quickly fix the order of the model and the number of calibration points required using results of this paper. The number of epochs needed to calibrate the sensor, and thereafter the epochs needed to recalibrate in event of sensitivity or offset drifts, are also presented to bring out the convergence time of the technique. More importantly, the advantages of the ANN technique over traditional regression based modeling are also discussed from the point of view of its advantage in hardware simplicity in microcontroller based implementation. Results presented in this paper would be of interest to instrumentation design engineers.

Algorithms↗
PubMed1994

Estimating marbling score in live cattle from ultrasound images using pattern recognition and neural network procedures.

Neural network processing of texture statistics (which parameterized longissimus muscle echograms of live cattle) resulted in marbling estimates that differed from corresponding USDA carcass marbling scores by an average of .42 marbling score units. This was more accurate (P < .001) than using the same features in a multiple regression model. Images were used from 53 cattle in the training set and from 108 cattle in the validation set. Over 500 texture statistics (including variations in direction, resolution, and step size) were screened to identify three candidates (Markovian homogeneity--step size = one; third quadrant emphasis from the bit-4, normalized run length/gray level matrix; and 12-pixel local standard deviation) for intensive analysis. The differences between the live animal estimates and carcass marbling were not much greater than the human error in assigning carcass marbling scores. When the results were subjected to receiver operating characteristic analysis, accuracies in grade classification were comparable to clinical, diagnostic imaging evaluations. It is feasible to incorporate this procedure into a computer interfaced with an ultrasound system to provide unsupervised instrument evaluation of live cattle in "near real time" (2 or 3 s).

Adipose Tissue↗
PubMed2002

Prediction of food thermal process evaluation parameters using neural networks.

Two neural networks (ANN) were developed to predict thermal process evaluation parameters g and f(h)/U (the ratio of heating rate index to the sterilizing value), respectively. The temperature change required for the thermal destruction curve to traverse one log cycle (z), cooling lag factor (j(c)) andf(h)/U were input variables for predicting g and z, while j(c) and g were inputs for predicting f(h)/U. The data used to train and verify the ANN were obtained from reported values. Shrinking of input and output variables using natural logarithm function improved the prediction accuracy. The use of "Wardnets" with three slabs of 14 nodes in each slab, with a learning rate of 0.7 and momentum of 0.9 provided the best predictions. The g (unshrunk values) was predicted with a mean relative error of 1.25 +/- 1.77%, and a mean absolute error of 0.11 +/- 0.16 degrees F. The f(h)/U was predicted with a mean relative error of 1.41 +/- 3.40%, and a mean absolute error of 2.43 +/- 15.97, using 10 nodes in each slab. The process time calculated using the g from the ANN models closely followed the time calculated from the tabulated gvalues (RMS=0.612 min, average absolute error=0.466 min with an S.D. of 0.400 min).

Algorithms↗
PubMed2005

Modeling of activity of cyclic urea HIV-1 protease inhibitors using regularized-artificial neural networks.

Artificial neural networks (ANNs) were used to model both inhibition of HIV-1 protease (K(i)) and inhibition of HIV replication (IC90) for 55 cyclic urea derivatives using constitutional and 2D descriptors. As a preliminary step, linear dependences were established by multiple linear regression (MLR) approaches, selecting the relevant descriptors by genetic algorithm (GA) feature selection. For ANN models non-linear GA feature selection was also applied. Non-linear modeling of K(i) overcame the results of the linear one using four properties, keeping in mind standard Pearson R correlation coefficients (0.931 vs. 0.862) and leave one out (LOO) cross-validation analysis (Q(LOO)2 = 0.703 vs. 0.510). On the other hand, IC90 modeling was insoluble by a linear approach: no predictive model was achieved; however, a non-linear relation was encountered according to statistic results (R = 0.891; Q(LOO)2 = 0.568). The best non-linear models suggested the influence of the presence of nitrogen atoms and the molecular volume distribution in the inhibitor structures on the HIV-1 protease inhibition as well as that the inhibition of HIV replication was dependent on the occurrence of five-member rings. Finally, inhibitors were well distributed regarding its activity levels in a Kohonen self-organizing map built using the input variables of the best non-linear models.

Algorithms↗
PubMed2005

Mapping the dose-effect relationship of orbofiban from sparse data with an artificial neural network.

A neural network (NN) pharmacodynamic model was developed that correlates the inhibition of ex vivo platelet aggregation by orbofiban, an oral glycoprotein IIb/IIIa antagonist, with the administered dose and patient characteristics. Data were obtained from a Phase-II dose-finding study of orbofiban in patients presenting with acute coronary syndromes. A back-propagation NN was designed to predict drug effect measured at predose and 4 and 6 h on treatment days 1, 28, and 84 (nine responses/patient). The training set consisted of patients for whom complete response profiles were reported (n = 67), and remaining patients were included in the validation data set (n = 47). The concentration-effect relationship was described additionally using a population direct-effect inhibitory sigmoidal model, and a comparison of the predictive performances of both models was performed. The final NN reasonably described orbofiban pharmacodynamics from sparse data sets without specifying a structural model or drug concentrations. Despite considerable inter-patient variability in response-time profiles, the population model revealed a strong correlation between drug concentration and effect and exhibited greater precision than the NN model. Although the population model showed greater precision, these results suggest that NNs may be useful for individualizing pharmacotherapy when drug concentrations are relatively unpredictable or unavailable.

Alanine↗
PubMedDate not supplied

Predicting conductance due to upconing using neural networks.

Artificial neural networks (ANNs) were developed to accurately predict highly time-variable specific conductance values in an unconfined coastal aquifer. Conductance values in the fresh water lens aquifer change in response to vertical displacements of the brackish zone and fresh water-salt water interface, which are caused by variable pumping and climate conditions. Unlike physical-based models, which require hydrologic parameter inputs, such as horizontal and vertical hydraulic conductivities, porosity, and fluid densities, ANNs can "learn" system behavior from easily measurable variables. In this study, the ANN input predictor variables were initial conductance, total precipitation, mean daily temperature, and total pumping extraction. The ANNs were used to predict salinity (specific conductance) at a single monitoring well located near a high-capacity municipal-supply well over time periods ranging from 30 d to several years. Model accuracy was compared against both measured/interpolated values and predictions were made with linear regression, and in general, excellent prediction accuracy was achieved. For example, although the average percent change of conductance over 90-d periods was 39%, the absolute mean prediction error achieved with the ANN was only 1.1%. The ANNs were also used to conduct a sensitivity analysis that quantified the importance of each of the four predictor variables on final conductance values, providing valuable insights into the dynamics of the system. The results demonstrate that the ANN technology can serve as a powerful and accurate prediction and management tool, minimizing degradation of ground water quality to the extent possible by identifying appropriate pumping policies under variable and/or changing climate conditions.

Environmental Monitoring↗
PubMed1994

Analysis of E.coli promoter structures using neural networks.

Backpropagation neural network is trained to identify E.coli promoters of all spacing classes (15 to 21). A three module approach is employed wherein the first neural net module predicts the consensus boxes, the second module aligns the promoters to a length of 65 bases and the third neural net module predicts the entire sequence of 65 bases taking care of the possible interdependencies between the bases in the promoters. The networks were trained with 106 promoters and random sequences which were 60% AT rich and tested on 126 promoters (Bacterial, Mutant and Phage promoters). The network was 98% successful in promoter recognition and 90.2% successful in non-promoter recognition when tested on 5000 randomly generated sequences. The network was further trained with 11 mutated non-promoters and 8 mutated promoters of the p22ant promoter. The testing set with 7 mutated promoters and 13 mutated non-promoters of p22ant were identified. The network was upgraded using total 1665 data of promoters and non-promoters to identify any promoter sequences in the gene sequences. The network identified the locations of P1, P2 and P3 promoters in the pBR322 plasmid. A search for the start codon, Ribosomal Binding Site and the stop codon by a string search procedure has also been added to find the possible promoters that can yield protein products. The network was also successfully tested on a synthetic plasmid pWM528.

Base Sequence↗