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 163 records · Page 9Linked to original sources

PubMed1998

Mapping of neural networks onto the memory-processor integrated architecture.

In this paper, an effective memory-processor integrated architecture, called memory-based processor array for artificial neural networks (MPAA), is proposed. The MPAA can be easily integrated into any host system via memory interface. Specifically, the MPA system provides an efficient mechanism for its local memory accesses allowed by row and column bases, using hybrid row and column decoding, which is suitable for computation models of ANNs such as the accessing and alignment patterns given for matrix-by-vector operations. Mapping algorithms to implement the multilayer perceptron with backpropagation learning on the MPAA system are also provided. The proposed algorithms support both neuron and layer level parallelisms which allow the MPAA system to operate the learning phase as well as the recall phase in the pipelined fashion. Performance evaluation is provided by detailed comparison in terms of two metrics such as the cost and number of computation steps. The results show that the performance of the proposed architecture and algorithms is superior to those of the previous approaches, such as one-dimensional single-instruction multiple data (SIMD) arrays, two-dimensional SIMD arrays, systolic ring structures, and hypercube machines.

Journal Article↗

PubMed1999

A novel learning algorithm which improves the partial fault tolerance of multilayer neural networks.

The paper deals with the problem of fault tolerance in a multilayer perceptron network. Although it already possesses a reasonable fault tolerance capability, it may be insufficient in particularly critical applications. Studies carried out by the authors have shown that the traditional backpropagation learning algorithm may entail the presence of a certain number of weights with a much higher absolute value than the others. Further studies have shown that faults in these weights is the main cause of deterioration in the performance of the neural network. In other words, the main cause of incorrect network functioning on the occurrence of a fault is the non-uniform distribution of absolute values of weights in each layer. The paper proposes a learning algorithm which updates the weights, distributing their absolute values as uniformly as possible in each layer. Tests performed on benchmark test sets have shown the considerable increase in fault tolerance obtainable with the proposed approach as compared with the traditional backpropagation algorithm and with some of the most efficient fault tolerance approaches to be found in literature.

Journal Article↗

PubMed2002

The effect of experimental design on the modeling of a tablet coating formulation using artificial neural networks.

The aim of this study was to investigate the effect of experimental design strategy on the modeling of a film coating formulation by artificial neural networks (ANNs). Box-Behnken, central composite and pseudo-random designs of 102, 90 and 100 simulated records, respectively were used to train a multilayer perceptron (MLP) ANN comprising six input and two output nodes separated by a single hidden layer of five nodes. Network over-training was limited by using a test set of 40 pseudo-randomly distributed records. The models were validated using a set of 60 pseudo-randomly distributed records. Crack velocity was highly curved with respect to pigment particle size and size distribution. Similarly, film opacity was highly curved in response to pigment concentration and film thickness. The Box-Behnken and central composite designs generated models that were unable to predict crack velocity and showed extensive bias in prediction of film opacity. The pseudo-random design was unable to predict crack velocity of the test data set but yielded acceptable predictions for the validation set. Film opacity was well predicted by the pseudo-random design model. The poor predictive ability of the Box-Behnken and central composite models was attributed to poor interpolation of the high curvature of the response surfaces. In contrast, the pseudo-random design mapped the interior of the design space allowing improved interpolation and predictive ability. It is concluded that Box-Behnken and central composite experimental designs are inappropriate for ANN modeling of highly curved responses and that extensive internal mapping of the design space is essential to generate predictive ANN models.

Drug Design↗

PubMed2003

Rule-base derivation for intensive care ventilator control using ANFIS.

In recent years, much research has been done on the use of fuzzy systems in medicine. The fuzzy rule-bases have usually been derived after extensive discussion with the clinical experts. This takes a lot of time from the clinical experts and the knowledge engineers. This paper presents the use of the adaptive neuro-fuzzy inference system (ANFIS) in rule-base derivation for ventilator control. The change of the inspired fraction of oxygen (FiO(2)) advised by eight clinical experts responding to 71 clinical scenarios was recorded. ANFIS and a multilayer perceptron (MLP) were then used to model the relationship between the inputs (the arterial oxygen tension (PaO(2)), FiO(2) and the positive end-expiratory pressure (PEEP) level) and the change in FiO(2) suggested. Compared to a previous fuzzy advisor (FAVeM), both the ANFIS and the MLP were found to correlate with the clinicians' decision better (correlation coefficient of 0.694 and 0.701, respectively compared to 0.630). A formerly developed model-based radial basis network advisor (RBN-MB) was used for comparison. Closed-loop simulations showed that the ANFIS, MLP and the RBN-MB's performance were comparable to the clinicians' performance (correlation coefficients of 0.852, 0.962 and 0.787, respectively). The FAVeM's performance differed from the clinicians' performance (correlation coefficient of 0.332) but the resulting PaO(2) was still within safety limits.

Computer Simulation↗

PubMed2001

Combinatorial networks.

A novel approach for the analysis and virtual screening of large combinatorial libraries is presented. The method attempts to relieve the computational burden by computing the properties of the products in a way that does not require their explicit enumeration. In particular, a small subset of compounds from the virtual library is identified and their descriptors are calculated in a conventional manner. The resulting data is used as input to a multilayer perceptron, which is trained to predict the descriptors of the products from the descriptors of their respective building blocks. Once trained, the neural network is able to estimate the descriptors of the remaining members of the virtual library with remarkable accuracy, without ever, generating their connection tables. This method eliminates the two most time-consuming steps in virtual screening and allows the processing of very large combinatorial libraries that are intractable with conventional techniques.

Chemistry Techniques, Analytical↗

PubMed2000

EEG-based discrimination between imagination of left and right hand movements using Adaptive Gaussian Representation.

This article uses the Adaptive Gaussian Representation (AGR) for human electroencephalogram (EEG) feature extraction aiming the discrimination among mental tasks to be used in a brain computer interface (BCI). It does not focus on the AGR time-frequency representation, but rather on their projection coefficients. Ten volunteers were asked to imagine either right or left hand movement, according to a proper visual stimulus. The features of the resulting EEG signals were characterised by extracting AGR coefficients. Classification was carried out using a Multilayer perceptron (MLP) trained with the classical backpropagation algorithm. Overall results show that AGR coefficients representation is able to reveal a significant EEG discrimination between imagination of right and left hand movement with a mean classification performance of 91%+/-5.8% achieved for female subjects and 87%+/-5.0% achieved for male subjects.

Adult↗

PubMed2004

Segmentation of magnetic resonance images using a combination of neural networks and active contour models.

Segmentation of medical images is very important for clinical research and diagnosis, leading to a requirement for robust automatic methods. This paper reports on the combined use of a neural network (a multilayer perceptron, MLP) and active contour model ('snake') to segment structures in magnetic resonance (MR) images. The perceptron is trained to produce a binary classification of each pixel as either a boundary or a non-boundary point. Subsequently, the resulting binary (edge-point) image forms the external energy function for a snake, used to link the candidate boundary points into a continuous, closed contour. We report here on the segmentation of the lungs from multiple MR slices of the torso; lung-specific constraints have been avoided to keep the technique as general as possible. In initial investigations, the inputs to the MLP were limited to normalised intensity values of the pixels from an (7 x 7) window scanned across the image. The use of spatial coordinates as additional inputs to the MLP is then shown to provide an improvement in segmentation performance as quantified using the effectiveness measure (a weighted product of precision and recall). Training sets were first developed using a lengthy iterative process. Thereafter, a novel cost function based on effectiveness is proposed for training that allows us to achieve dramatic improvements in segmentation performance, as well as faster, non-iterative selection of training examples. The classifications produced using this cost function were sufficiently good that the binary image produced by the MLP could be post-processed using an active contour model to provide an accurate segmentation of the lungs from the multiple slices in almost all cases, including unseen slices and subjects.

Algorithms↗

PubMed2000

Improving diagnostic accuracy using a hierarchical neural network to model decision subtasks.

A number of quantitative models including linear discriminant analysis, logistic regression, k nearest neighbor, kernel density, recursive partitioning, and neural networks are being used in medical diagnostic support systems to assist human decision-makers in disease diagnosis. This research investigates the decision accuracy of neural network models for the differential diagnosis of six erythematous-squamous diseases. Conditions where a hierarchical neural network model can increase diagnostic accuracy by partitioning the decision domain into subtasks that are easier to learn are specifically addressed. Self-organizing maps (SOM) are used to portray the 34 feature variables in a two dimensional plot that maintains topological ordering. The SOM identifies five inconsistent cases that are likely sources of error for the quantitative decision models; the lower bound for the diagnostic decision error based on five errors is 0.0140. The traditional application of the quantitative models cited above results in diagnostic error levels substantially greater than this target level. A two-stage hierarchical neural network is designed by combining a multilayer perceptron first stage and a mixture-of-experts second stage. The second stage mixture-of-experts neural network learns a subtask of the diagnostic decision, the discrimination between seborrheic dermatitis and pityriasis rosea. The diagnostic accuracy of the two stage neural network approaches the target performance established from the SOM with an error rate of 0.0159.

Chronic Disease↗

PubMed1990

Spectral peak verification and recognition using a multilayered neural network.

The verification and recognition of peak-shaped signals in analytical data are ubiquitous scientific problems. Experimental data contain overlapping signals and noise, which make sensitive and reliable peak recognition difficult. A peak detection system based on a class of neural networks known as "multilayered perceptrons" has been created. The network was trained and evaluated with use of vapor-phase infrared spectral data. The results of varying the network architecture on system training and prediction performance along with refinement of the form of the input pattern are presented.

Artificial Intelligence↗

PubMedDate not supplied

Artificial neural networks and linear discriminant analysis: a valuable combination in the selection of new antibacterial compounds.

A set of topological descriptors has been used to discriminate between antibacterial and nonantibacterial drugs. Topological descriptors are simple integers calculated from the molecular structure represented in SMILES format. The methods used for antibacterial activity discrimination were linear discriminant analysis (LDA) and artificial neural networks of a multilayer perceptron (MLP) type. The following plot frequency distribution diagrams were used: a function of the number of drugs within a value interval of the discriminant function and the output value of the neural network versus these values. Pharmacological distribution diagrams (PDD) were used as a visualizing technique for the identification of antibacterial agents. The results confirmed the discriminative capacity of the topological descriptors proposed. The combined use of LDA and MLP in the guided search and the selection of new structures with theoretical antibacterial activity proved highly effective, as shown by the in vitro activity and toxicity assays conducted.

Anti-Bacterial Agents↗

PubMedDate not supplied

Elucidation of characteristic structural features of ligand binding sites of protein kinases: a neural network approach.

Protein kinases play important roles in regulating cellular signal transduction and other biochemical processes, and they are attractive targets for drug discovery programs in many disease areas. Most kinase inhibitors under development as drugs act by directly competing with ATP at the ATP-binding site of the kinase. There are more than 500 protein kinases, and the ATP-binding site is highly conserved among them. Therefore selectivity is an essential requirement for clinically effective drugs, and understanding the structural characteristics of ATP-binding sites is of crucial importance. The objective of the present study was to elucidate the structural characteristics of the adenosine-binding site of four major kinase groups, AGC (PKA, PKG, and PKC families), CaMK (calcium/calmodulin-dependent protein kinases), CMGC (CDK, MAPK, GSK3, and CLK families), and TK (tyrosine kinases). To do this, we classified the kinases into groups by using feed-forward multilayer perceptron (MLP) neural networks and structural, electronic, and hydrophobic descriptors of the amino acids at the adenosine-binding site. A total of 275 kinases were classified in two ways: (1) kinases belonging to a certain group were distinguished from those not belonging to that group, and (2) all of the kinases were classified into four groups. More than 85% of the kinases were correctly classified by both methods. Trained neural networks clarified which amino acids and which properties characterize the adenosine-binding site of each group, and the results were visualized by molecular graphics. Comparison of the modeled neural networks and the distributions of amino acids provided more detailed information on the structural characteristics of each group. Application of the present results to drug development is also discussed.

Adenosine Triphosphate↗

PubMed2002

Aroma quality differentiation of pyrazine derivatives using self-organizing molecular field analysis and artificial neural network.

The encoding of various aroma impressions and the distinction between different aroma qualities are unsolved problems, as differences between aroma impressions can be described only in a qualitative but not in a quantitative manner. As a consequence, classifications of various aroma qualities cannot easily be performed by standard QSAR methods. To find a proper way to encode aroma impressions for SAR studies, a total of 50 pyrazine-based aroma compounds showing the aroma quality of earthy, green-earthy, or green are analyzed. Special attention is thereby turned on the mixed aroma impression green-earthy. Classifications on the whole data set as well as on smaller subsets are calculated using self-organizing molecular field analysis (SOMFA) and artificial neural networks (ANNs). SOMFA classifies between two or three aroma impressions, leading to models satisfying in predictive power. ANN analysis using multilayer perceptron network architecture with one hidden layer and nominal output as well as genetic regression neural network) with two hidden layers and numerical output both lead to a rather good performance rate of 94%.

Neural Networks, Computer↗

PubMed2006

Supervised pattern recognition procedures for discrimination of whiskeys from gas chromatography/mass spectrometry congener analysis.

The volatile congener analysis of 52 commercialized whiskeys (24 samples of single malt Scotch whiskey, 18 samples of bourbon whiskey, and 10 samples of Irish whiskey) was carried out by gas chromatography/mass spectrometry after liquid-liquid extraction with dichloromethane. Pattern recognition procedures were applied for discrimination of different whiskey categories. Multivariate data analysis includes linear discriminant analysis (LDA), k nearest neighbors (KNN), soft independent modeling of class analogy (SIMCA), procrustes discriminant analysis (PDA), and artificial neural networks techniques involving multilayer perceptrons (MLP) and probabilistic neural networks (PNN). Classification rules were validated by considering the number of false positives (FPs) and false negatives (FNs) of each class associated to the prediction set. Artificial neural networks led to the best results because of their intrinsic nonlinear features. Both techniques, MLP and PNN, gave zero FPs and zero FNs for all of the categories. KNN is a nonparametric method that also provides zero FPs and FNs for every class but only when selecting K = 3 neighbors. PDA produced good results also (zero FPs and FNs always) but only by selecting nine principal components for class modeling. LDA shows a lesser classification performance, because of the building of linear frontiers between classes that does not apply in many real situations. LDA led to one FP for bourbons and one FN for scotches. The worse results were obtained with SIMCA, which gave a higher number of FPs (five for both scotches and bourbons) and FNs (six for scotchs and two for bourbons). The possible cause of these findings is the strong influence of class inhomogeneities on the SIMCA performance. It is remarkable that in any case, all of the methodologies lead to zero FPs and FNs for the Irish whiskeys.

Alcoholic Beverages↗

PubMed2001

Prediction of the aroma quality and the threshold values of some pyrazines using artificial neural networks.

An artificial neural network is used to predict both the classification of aroma compounds and their flavor impression threshold values for a series of pyrazines. The classification set consists of 98 compounds (32 green, 43 bell-pepper, and 23 nutty smelling pyrazines), and the regression sets consist of 24 green and 37 bell-pepper odorous pyrazines. The best classification of the three aroma impressions (93.7%) is obtained by using a multilayer perceptron network architecture. To predict the threshold values of bell-pepper fragrance, a standard Pearson R correlation coefficient of 0.936 for the training set, 0.912 for the verification set, and 0.926 for the test set is received with two hidden layers consisting of two and one neurons. The network for the threshold prediction of the class of green-smelling pyrazines with one hidden layer containing three neurons turns out to be the best with a standard Pearson R correlation coefficient of 0.859 for the training, 0.918 for the verification, and 0.948 for the test set. These good correlations show that artificial neural networks are versatile tools for the classification of aroma compounds.

Flavoring Agents↗

PubMed2001

Metabonomic characterization of genetic variations in toxicological and metabolic responses using probabilistic neural networks.

Current emphasis on efficient screening of novel therapeutic agents in toxicological studies has resulted in the evaluation of novel analytical technologies, including genomic (transcriptomic) and proteomic approaches. We have shown that high-resolution 1H NMR spectroscopy of biofluids and tissues coupled with appropriate chemometric analysis can also provide complementary data for use in in vivo toxicological screening of drugs. Metabonomics concerns the quantitative analysis of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification [Nicholson, J. K., Lindon, J. C., and Holmes, E. (1999) Xenobiotica 11, 1181-1189]. In this study, we have used 1H NMR spectroscopy to characterize the time-related changes in the urinary metabolite profiles of laboratory rats treated with 13 model toxins and drugs which predominantly target liver or kidney. These 1H NMR spectra were data-reduced and subsequently analyzed using a probabilistic neural network (PNN) approach. The methods encompassed a database of 1310 samples, of which 583 comprised a training set for the neural network, with the remaining 727 (independent cases) employed as a test set for validation. Using these techniques, the 13 classes of toxicity, together with the variations associated with strain, were distinguishable to >90%. Analysis of the 1H NMR spectral data by multilayer perceptron networks and principal components analysis gave a similar but less accurate classification than PNN analysis. This study has highlighted the value of probabilistic neural networks in developing accurate NMR-based metabonomic models for the prediction of xenobiotic-induced toxicity in experimental animals and indicates possible future uses in accelerated drug discovery programs. Furthermore, the sensitivity of this tool to strain differences may prove to be useful in investigating the genetic variation of metabolic responses and for assessing the validity of specific animal models.

Animals↗

PubMed2000

Autoassociative MLP in sleep spindle detection.

Spindles are one of the most important short-lasting waveforms in sleep EEG. They are the hallmarks of the so-called Stage 2 sleep. Visual spindle scoring is a tedious workload, since there are often a thousand spindles in one all-night recording of some 8 hr. Automated methods for spindle detection typically use some form of fixed spindle amplitude threshold, which is poor with respect to inter-subject variability. In this work a spindle detection system allowing spindle detection without an amplitude threshold was developed. This system can be used for automatic decision making of whether or not a sleep spindle is present in the EEG at a certain point of time. An Autoassociative Multilayer Perceptron (A-MLP) network was employed for the decision making. A novel training procedure was developed to remove inconsistencies from the training data, which was found to improve the system performance significantly.

Adult↗

PubMed2004

Classification of mitral insufficiency and stenosis using MLP neural network and neuro-fuzzy system.

Cardiac Doppler signals recorded from mitral valve of 60 patients were transferred to a personal computer by using a 16-bit sound card. The power spectral density (PSD) was applied to the recorded signal from each patient. In order to do a good interpretation and rapid diagnosis, PSD values classified using multilayer perceptron (MLP) and neuro-fuzzy system. Our findings demonstrated that 93.33% classification success rate was obtained from MLP, 90% classification success rate was obtained from neuro-fuzzy system. The classification results show that MLP offers best results in the case of diagnosis.

Diagnosis, Computer-Assisted↗

PubMed2004

Classification of MCA stenosis in diabetes by MLP and RBF neural network.

For the classification of Middle Cerebral Artery (MCA) stenosis, Doppler signals have been received from the diabetes and control group by using 2 MHz Transcranial Doppler. After the Fast Fourier Transform (FFT) analyses of the Doppler signals, Power Spectrum Density (PSD) estimations have been made and Multilayer Perceptron (MLP) and Radial Basis Function (RBF) have been dealt to apply to the neural networks. PSD estimations of Doppler signals received from MCA of 104 subjects have been successfully classified by MLP (correct classification = 94.2%) and RBF (correct classification = 88.4%) neural network. As we have seen in the area under ROC curve (AUC), MLP neural network (AUC = 0.934) has classified more successfully when compared with RBF neural network (AUC = 0.873).

Brain↗