PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Neural Networks, Computer”

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 1,315 records · Page 73Linked to original sources
PubMed2007

Exact subthreshold integration with continuous spike times in discrete-time neural network simulations.

Very large networks of spiking neurons can be simulated efficiently in parallel under the constraint that spike times are bound to an equidistant time grid. Within this scheme, the subthreshold dynamics of a wide class of integrate-and-fire-type neuron models can be integrated exactly from one grid point to the next. However, the loss in accuracy caused by restricting spike times to the grid can have undesirable consequences, which has led to interest in interpolating spike times between the grid points to retrieve an adequate representation of network dynamics. We demonstrate that the exact integration scheme can be combined naturally with off-grid spike events found by interpolation. We show that by exploiting the existence of a minimal synaptic propagation delay, the need for a central event queue is removed, so that the precision of event-driven simulation on the level of single neurons is combined with the efficiency of time-driven global scheduling. Further, for neuron models with linear subthreshold dynamics, even local event queuing can be avoided, resulting in much greater efficiency on the single-neuron level. These ideas are exemplified by two implementations of a widely used neuron model. We present a measure for the efficiency of network simulations in terms of their integration error and show that for a wide range of input spike rates, the novel techniques we present are both more accurate and faster than standard techniques.

Action Potentials↗
PubMed1995

A simple competitive account of some response properties of visual neurons in area MSTd.

A simple and biologically plausible model is proposed to simulate the optic flow computation taking place in the dorsal part of medial superior temporal (MSTd) area of the visual cortex in the primates' brain. The model is a neural network composed of competitive learning layers. The input layer of the network simulates the neurons in the middle temporal (MT) area that selectively respond to the visual stimuli of the input motion patterns with different local velocities. The output layer of the network simulates the MSTd neurons that selectively respond to different types of optic flow motion patterns including planar, circular, radial, and spiral motions. Simulation results obtained from this model show that the behaviors of the output nodes of the network resemble very closely the known responsive properties of the MSTd neurons found neurophysiologically, such as the existence of three types of MSTd neurons that respond, respectively, to one, two, or three types of the input motion patterns with different position dependences, and the continuum of response selectivity formed by the three types of neurons.

Animals↗
PubMed2000

Modeling splice sites with Bayes networks.

MOTIVATION: The main goal in this paper is to develop accurate probabilistic models for important functional regions in DNA sequences (e.g. splice junctions that signal the beginning and end of transcription in human DNA). These methods can subsequently be utilized to improve the performance of gene-finding systems. The models built here attempt to model long-distance dependencies between non-adjacent bases. RESULTS: An efficient modeling method is described which models biological data more accurately than a first-order Markov model without increasing the number of parameters. Intuitively, a small number of parameters helps a learning system to avoid overfitting. Several experiments with the model are presented, which show a small improvement in the average accuracy as compared with a simple Markov model. These experiments suggest that single long distance dependencies do not help the recognition problem, thus confirming several previous studies which have used more heuristic modeling techniques. AVAILABILITY: This software is available for downloaded and as a web resource at http://www.ai.uic.edu/software CONTACT: kasif@eecs.uic.edu

Bayes Theorem↗
PubMed2003

A realistic neural-network simulation of both slow and quick phase components of the guinea pig VOR.

A realistic neural-network model was constructed to simulate production of both the slow-phase and quick-phase components of vestibular nystagmus by incorporating a quick-phase pathway into a previous model of the slow phase. The neurons in the network were modelled by multicompartmental Hodgkin-Huxley-style spiking neurons based on known responses and projections of physiologically identified vestibular neurons. The modelling used the GENESIS software package. The slow-phase network consisted of ganglion and medial vestibular nucleus (MVN) neurons; the latter were constructed using biophysical models of MVN type A and B neurons. The quick-phase network contained several types of bursting cells which have been shown to have major roles in the generation of the quick phase: burster-driver neurons, long-lead burst neurons, pause neurons, excitatory burst neurons and inhibitory burst neurons. Comparison of the output neural responses from the model with guinea pig behavioural responses from the companion paper showed consistency between model and animal data for neuron firing patterns, maximal firing rates, and timing, duration and number of quick phases. Comparisons were made for stable head input and for sinusoidal angular stimuli at a range of frequencies from 0.1 to 2 Hz. Except for data at 0.1 Hz, where the simulation produced one more quick phase per half cycle than the animal data, the number of quick phases was consistent between the model and the animal data. The model was also used to simulate the effects both of unilateral vestibular deafferentation (UVD) and of vestibular compensation after UVD, and the responses in the modelled MVN neurons were affected in a way similar to those measured in guinea pig MVN neurons: the number of quick phases and their timing changed in a similar fashion to that observed in behavioural data.

Animals↗
PubMed2000

Correctness of local probability in graphical models with loops.

Graphical models, such as Bayesian networks and Markov networks, represent joint distributions over a set of variables by means of a graph. When the graph is singly connected, local propagation rules of the sort proposed by Pearl (1988) are guaranteed to converge to the correct posterior probabilities. Recently a number of researchers have empirically demonstrated good performance of these same local propagation schemes on graphs with loops, but a theoretical understanding of this performance has yet to be achieved. For graphical models with a single loop, we derive an analytical relationship between the probabilities computed using local propagation and the correct marginals. Using this relationship we show a category of graphical models with loops for which local propagation gives rise to provably optimal maximum a posteriori assignments (although the computed marginals will be incorrect). We also show how nodes can use local information in the messages they receive in order to correct their computed marginals. We discuss how these results can be extended to graphical models with multiple loops and show simulation results suggesting that some properties of propagation on single-loop graphs may hold for a larger class of graphs. Specifically we discuss the implication of our results for understanding a class of recently proposed error-correcting codes known as turbo codes.

Bayes Theorem↗
PubMed2000

[The use of artificial neuronal networks in the treatment of peptic ulcer].

Today the use of artificial neuronic nets (ANN) is the most perspective method in the prediction of clinical course and in the choice of correct therapeutic tactics. The experimental complex "Axon" with the software was created in the hospital that permitted to predict the course if gastric and duodenal ulcers. We introduced the data of 909 patients that contained the anamnesis, results of objective examination and late results of the treatment. These data has become the base for training and formation of ANN similar to the structure of human cerebral cortex. Having received the information about a new patient and using the previous experience of therapy of 909 patients mentioned above the trained ANN gives an accurate information concerning this patient. It is presented in the form of protocol of prognosis of ulcer clinical course, the probability of complication development, the results of conservative and surgical treatment. The protocol includes the conclusion with recommendations on the choice of therapeutic tactics.

Chronic Disease↗
PubMedDate not supplied

Neuronal network simulating the simultaneous contrast of equiluminant colors.

A vector model for the simultaneous contrast of equiluminant colors is proposed. The main characteristics of simultaneous color contrast are associated with the interaction of the test and induction fields. The mechanisms of color contrast depend on the activity of neurons which subtract excitation arising in analogous neurons of the test and induction fields. Signals from the "subtractor" neurons pass via local inputs to "adder" neurons. Selective color detectors are excited by contributions from the adder neurons. When the induction and test fields are of identical color and their excitation vectors coincide, the color of the test field depends on neurons excited only in the test field, and the induction field has no effect on the color of the test field. When the test and induction fields are of different color tone or saturation, the color of the induction field affects the color of the test field such that the excitation vector of the test field diverges from the excitation vector of the induction field. This model can be used for analysis of neurophysiological and psychophysiological studies of color contrast.

Color Perception↗
PubMed2003

A computational model of prefrontal control in free recall: strategic memory use in the California Verbal Learning Task.

Several decades of research into the function of the frontal lobes in brain-damaged patients, and more recently in intact individuals using function brain imaging, has delineated the complex executive functions of the frontal cortex. And yet, the mechanisms by which the brain achieves these functions remain poorly understood. Here, we present a computational model of the role of the prefrontal cortex (PFC) in controlled memory use that may help to shed light on the mechanisms underlying one aspect of frontal control: the development and deployment of recall strategies. The model accounts for interactions between the PFC and medial temporal lobe in strategic memory use. The PFC self-organizes its own mnemonic codes using internally derived performance measures. These mnemonic codes serve as retrieval cues by biasing retrieval in the medial temporal lobe memory system. We present data from three simulation experiments that demonstrate strategic encoding and retrieval in the free recall of categorized lists of words. Experiment 1 compares the performance of the model with two control networks to evaluate the contribution of various components of the model. Experiment 2 compares the performance of normal and frontally lesioned models to data from several studies using frontally intact and frontally lesioned individuals, as well as normal, healthy individuals under conditions of divided attention. Experiment 3 compares the model's performance on the recall of blocked and unblocked categorized lists of words to data from Stuss et al. (1994) for individuals with control and frontal lobe lesions. Overall, our model captures a number of aspects of human performance on free recall tasks: an increase in total words recalled and in semantic clustering scores across trials, superiority on blocked lists of related items compared to unblocked lists of related items, and similar patterns of performance across trials in the normal and frontally lesioned models, with poorer overall performance of the lesioned models on all measures. The model also has a number of shortcomings, in light of which we suggest extensions to the model that would enable more sophisticated forms of strategic control.

Brain Injuries↗
PubMed1993

Cellular automata approaches to biological modeling.

We review a number of biologically motivated cellular automata (CA) that arise in models of excitable and oscillatory media, in developmental biology, in neurobiology, and in population biology. We suggest technical and theoretical arguments that permit greater speed and enhanced realism, and apply these to several classical examples of pattern formation. We also describe CA that arise in models for fibroblast aggregation, branching networks, trail following, and neuronal maps.

Allergy and Immunology↗
PubMed2006

Generalized RLS approach to the training of neural networks.

Recursive least square (RLS) is an efficient approach to neural network training. However, in the classical RLS algorithm, there is no explicit decay in the energy function. This will lead to an unsatisfactory generalization ability for the trained networks. In this paper, we propose a generalized RLS (GRLS) model which includes a general decay term in the energy function for the training of feedforward neural networks. In particular, four different weight decay functions, namely, the quadratic weight decay, the constant weight decay and the newly proposed multimodal and quartic weight decay are discussed. By using the GRLS approach, not only the generalization ability of the trained networks is significantly improved but more unnecessary weights are pruned to obtain a compact network. Furthermore, the computational complexity of the GRLS remains the same as that of the standard RLS algorithm. The advantages and tradeoffs of using different decay functions are analyzed and then demonstrated with examples. Simulation results show that our approach is able to meet the design goals: improving the generalization ability of the trained network while getting a compact network.

Algorithms↗
PubMed1997

The role of the Internet in medical decision making.

The Internet has dramatically changed the means by which information is obtained. Accurate, up-to-date information is vital to maintain the quality of healthcare, especially as US healthcare delivery changes to a primary care-based system. The availability of this new and potentially vast source of information also affects strategies for medical decision making. In this article, use of online information in our medical center is discussed, together with the impact of a locally-developed decision support system. This system first contained components for differential diagnosis as well as computer-assisted instruction. Initially, online searching was limited to Medline literature searches. This component has now been expanded to include important new tools for accessing medical information on the Internet.

Artificial Intelligence↗
PubMedDate not supplied

The effect of data sampling on the performance evaluation of artificial neural networks in medical diagnosis.

PURPOSE: To study the effect of data sampling on the predictive assessment of artificial neural networks (ANNs) for medical diagnostic tasks. METHODS: Three statistical techniques were used to evaluate the diagnostic performances of ANNs: 1) cross validation, 2) round robin, and 3) bootstrap. These techniques are different sampling plans designed to reduce the small-sample estimation bias and variance contributions. The study was based on two networks, one developed for the diagnosis of pulmonary embolism (1,064 cases) and the other developed for the diagnosis of breast cancer (206 cases). RESULTS: The three sampling techniques produced different performance estimates for both networks. The estimates varied substantially depending on the training sample size and the training-stopping criterion. CONCLUSION: The predictive assessment of ANNs in medical diagnosis can vary substantially based on the complexity of the problem, the data sampling technique, and the number of cases available.

Breast Neoplasms↗
PubMed2004

Protein structure and fold prediction using tree-augmented naive Bayesian classifier.

For determining the structure class and fold class of Protein Structure, computer-based techniques have became essential considering the large volume of the data. Several techniques based on sequence similarity. Neural Networks, SVMs, etc have been applied. This paper presents a framework using the Tree-Augmented Networks (TAN) based on the theory of learning Bayesian networks but with less restrictive assumptions than the naive Bayesian networks. In order to enhance TAN's performance, pre-processing of data is done by feature discretization and post-processing is done by using Mean Probability Voting (MPV) scheme. The advantage of using Bayesian approach over other learning methods is that the network structure is intuitive. In addition, one can read off the TAN structure probabilities to determine the significance of each feature (say, Hydrophobicity) for each class, which help to further understand the mystery of protein structure. Experimental results and comparison with other works over two databases show the effectiveness of our TAN based framework. The idea is implemented as the BAYESPROT web server and it is available at http://www-appn.comp.nus.edu.sg/-bioinfo/bayesprot/Default.htm.

Artificial Intelligence↗
PubMed1996

The application of an artificial neural network to Doppler ultrasound waveforms for the classification of arterial disease.

In this study we have investigated the application of an Artificial Neural Net classifier to the diagnosis of vascular disease using Doppler ultrasound blood-velocity/time waveforms. A multi-layer perceptron network was trained with waveforms from control subjects and from patients with arterial disease. The diseased cases were confirmed by angiography and allocated to three groups according to the location of the stenosis: proximal or distal to the site of measurement or multi-segmental. We compared network classification results with a Bayesian classifier following a Principal Component Analysis of the waveforms. Versions of both classifiers were trained to discriminate two classes (normal v. abnormal) and four classes. In both cases the neural networks gave superior discrimination to the Bayesian classifier. While the four-class network was unable to provide useful discrimination among the stenosis sites, discrimination between abnormal classes was obtained which is comparable to that achieved by a human expert observer.

Adult↗
PubMed1993

Hydra: a C-language environment for real-time DOS multitasking at the bedside.

Patient monitoring at the bedside is an inherently parallel job, best handled by multiple individual tasks running concurrently. Cost and diffusion considerations strongly favor the use of PC's at the bedside, but their most widespread operating system, DOS, is not built for multitasking. Hence, a software platform in C language has been prepared, allowing the intermediate programmer to easily write independent modules which will then run simultaneously without conflicts. Such a platform aims at allowing effortless sharing of data among concurrently running processes, while providing strong insulation between tasks, enough to allow multiple copies of any one task to run simultaneously unknown to each other. A cooperative, memory sharing multitasking paradigm has been chosen, which offers fine granularity of timeslicing and low execution overhead at the price of some loss in generality of design. Speed, data exchange capability and number of stackable windows are greater than with commercial packages like Windows or LabWindows. Dynamical reprioritization of tasks is built in, allowing the computerized monitor to focus its attention and resources on urgent tasks.

Algorithms↗
PubMed2003

Predicting cranial computed tomography results of head injury patients using an artificial neural network.

We conducted this study to determine if artificial neural network modeling would predict cranial computed tomography results in head injury patients using different combinations of clinical variables. 150 consecutive patients admitted to a regional trauma center with head injury were enrolled in the study. The CT was labeled with presence of surgically significant intracranial hematoma (SSIH), if midline shift, obliteration of ambient cistern or basal cistern were found. The best performance of our models to differentiate normal from abnormal cranial CT and detection of SSIH was ideal.

Craniocerebral Trauma↗
PubMed1998

A neural model of binocular integration and rivalry based on the coordination of action-potential timing in primary visual cortex.

In normal vision, the inputs from the two eyes are integrated into a single percept. When dissimilar images are presented to the two eyes, however, they compete for perceptual dominance, so that one eye's view suppresses that of the other. Recent evidence suggests that this phenomenon, known as binocular rivalry, arises through competition between alternative stimulus interpretations in extrastriate cortex. Because eye-specific information appears to be lost at this stage, it remains unclear how the stimulus conditions that yield binocular rivalry are distinguished from those that produce stable single vision. Using a neural network that models the mammalian early visual system, I investigate here the hypothesis that congruent and conflicting stimuli are distinguished by their different effects on the relative timing of action potentials in primary visual cortex (V1), where monocular inputs are first combined. In the model, congruent stimulation of both eyes results in synchronization of discharges among binocular neurons in V1. By contrast, conflicting stimulation of the two eyes results in neuronal asynchrony in this area. This asynchrony then produces rivalrous response suppression at later stages in the visual pathway. Synchronization of firing in V1, however, prevents such competition, thereby ensuring non-rivalrous responses. These novel effects of spike timing on competition emerge naturally from the network dynamics. The results suggest that input-related differences in relative spike timing at an early stage of visual processing may play an important part in the phenomena both of binocular integration and rivalry; furthermore, they indicate that the temporal patterning of cortical activity may be a fundamental mechanism of selection among competing stimulus representations.

Action Potentials↗