Discrete Neural Network Research Articles

Discrete neural network ART using a Hamming distance

Disadvantages of image proximity measure that is used in discrete neural network of adaptive resonance theory (ART) ART-1 are analyzed. The discrete network ART that uses the Hamming distance as an image proximity measure is proposed. Stable plastic Hamming neural network that is able to recognize the new information on its inputs is developed. Figs.: 2. Refs.: 9 titles. Keywords: neural network; adaptive resonance theory; Hamming distance; image proximity measure.

A novel summation inequality for stability analysis of discrete-time neural networks

In this paper, based on the discrete Wirtinger inequality, a novel summation inequality is established which extends the Jensen inequality. By the technique of the novel inequality, a sufficient criterion on asymptotic stability of discrete neural networks with time-varying delays is obtained. A numerical example is given to demonstrate the effectiveness of the inequality.

Assessment of Eating Behavior Disturbance and Associated Neural Networks in Frontotemporal Dementia.

Abnormal eating behaviors are common in patients with frontotemporal dementia (FTD), yet their exact prevalence, severity, and underlying biological mechanisms are not understood. To define the severity of abnormal eating behavior and sucrose preference and their neural correlates in patients with behavioral variant FTD (bvFTD) and semantic dementia. Forty-nine patients with dementia (19 with bvFTD, 15 with semantic dementia, and 15 with Alzheimer disease) were recruited, and their eating behavior was compared with that of 25 healthy controls. The study was conducted from November 1, 2013, through May 31, 2015, and data analyzed from June 1 to August 31, 2015. Patients participated in an ad libitum breakfast test meal, and their total caloric intake and food preferences were measured. Changes in eating behavior were also measured using the Appetite and Eating Habits Questionnaire (APEHQ) and the Cambridge Behavioral Inventory (CBI). Sucrose preference was tested by measuring liking ratings of 3 desserts of varying sucrose content (A: 26%, B: 39%, C: 60%). Voxel-based morphometry analysis of whole-brain 3-T high-resolution brain magnetic resonance imaging was used to determine the gray matter density changes across groups and their relations to eating behaviors. Mean (SD) ages of patients in all 4 groups ranged from 62 (8.3) to 66 (8.4) years. At the ad libitum breakfast test meal, all patients with bvFTD had increased total caloric intake (mean, 1344 calories) compared with the Alzheimer disease (mean, 710 calories), semantic dementia (mean, 573 calories), and control groups (mean, 603 calories) (P < .001). Patients with bvFTD and semantic dementia had a strong sucrose preference compared with the other groups. Increased caloric intake correlated with atrophy in discrete neural networks that differed between patients with bvFTD and semantic dementia but included the cingulate cortices, thalami, and cerebellum in patients with bvFTD, with the addition of the orbitofrontal cortices and nucleus accumbens in patients with semantic dementia. A distributed network of neural correlates was associated with sucrose preference in patients with FTD. Marked hyperphagia is restricted to bvFTD, present in all patients with this diagnosis, and supports its diagnostic value. Differing neural networks control eating behavior in patients with bvFTD and semantic dementia and are likely responsible for the differences seen, with a similar network controlling sucrose preference. These networks share structures that control cognitive-reward, autonomic, neuroendocrine, and visual modulation of eating behavior. Delineating the neural networks involved in mediating these changes in eating behavior may enable treatment of these features in patients with complex medical needs and aid in our understanding of structures that control eating behavior in patients with FTD and healthy individuals.

$$H_{\infty }$$ H ∞ filtering for discrete-time fuzzy stochastic neural networks with mixed time-delays

The $$H_{\infty }$$ filter problem for a class of fuzzy stochastic discrete neural networks system with mixed delays is studied in this paper. The mixed delays consist of discrete and distributed delays. Based on discrete inequality technic and the Lyapunov–Krasovskii functional approach, sufficient conditions for the existence of admissible filters are established in terms of linear matrix inequalities, which ensure the asymptotical mean-square stability as well as a prescribed $$H_{\infty }$$ disturbance attenuation level. Examples and simulations are provided to illustrate the effectiveness of the proposed method.

A Max-Sum algorithm for training discrete neural networks

.We present an efficient learning algorithm for the problem of training neural networks with discrete synapses, a well-known hard (NP-complete) discrete optimization problem. The algorithm is a variant of the so-called Max-Sum (MS) algorithm. In particular, we show how, for bounded integer weights with q distinct states and independent concave a priori distribution (e.g. l1 regularization), the algorithm’s time complexity can be made to scale as per node update, thus putting it on par with alternative schemes, such as Belief Propagation (BP), without resorting to approximations. Two special cases are of particular interest: binary synapses and ternary synapses with l0 regularization. The algorithm we present performs as well as BP on binary perceptron learning problems, and may be better suited to address the problem on fully-connected two-layer networks, since inherent symmetries in two layer networks are naturally broken using the MS approach.

Image Restoration Technology Based on Discrete Neural network

With the development of computer science and technology, the development of artificial intelli- gence advances rapidly in the field of image restoration. Based on the MATLAB platform, this paper constructs a kind of image restoration technology of artificial intelligence based on the discrete neural network and feedfor- ward network, and carries out simulation and contrast of the restoration process by the use of the bionic algo- rithm. Through the application of simulation restoration technology, this paper verifies that the discrete neural network has a good convergence and identification capability in the image restoration technology with a better effect than that of the feedforward network. The restoration technology based on the discrete neural network can provide a reliable mathematical model for this field.

The Design and Implementation of Key Technologies of the Computer Algorithm Dynamic System

This topic mainly aimed at the dynamic simulation of some actual problems. The research is suitable for the dynamic system of modeling artificial model and learning algorithm. First it summarizes the existing network simulation model of dynamic system and network delay unit. All the feedback network and part of the feedback network model, the network topology is given and used for system simulation. The method of simulation learning algorithm and simulation instance. And then established the time-varying input and output process neural networks and network model of two kinds of discrete process neural network for its properties are analyzed and proved that the concrete learning algorithm is deduced. Finally combined with the actual problem, time-varying input and output process neural network is given in the application of system identification.

ARCHITECTURE AND ALGORITHMS OF NEURAL NETWORKS HAMMING AND HEBB, CAPABLE LEARN AND IDENTIFY NEW INFORMATION

The problem of the classical discrete neural networks Hamming and Hebb lossless previously stored information additional training. The object of research is the process of recognition and classification of images on systems that are based on artificial neural networks. The subject of research is the architecture and algorithms of artificial neural networks. Objective: To develop a stable and plastic neural networks Hamming and Hebb. The architecture and algorithms of discrete stable and plastic neural networks Hamming and Hebb, which not only can be trained during functioning, but also to recognize the new information. New networks can be an alternative to discreteneural network adaptive resonance theory. The developed approach for training can be generalized to other neural networks. Experimental investigations of the functioning of the developed algorithms of artificial neural networks. The experimental results confirm the validity of the proposed approach.

Neural networks art: solving problems with multiple solutions and new teaching algorithm.

A new discrete neural networks adaptive resonance theory (ART), which allows solving problems with multiple solutions, is developed. New algorithms neural networks teaching ART to prevent degradation and reproduction classes at training noisy input data is developed. Proposed learning algorithms discrete ART networks, allowing obtaining different classification methods of input.

Neuroendocrine control of seasonal plasticity in the auditory and vocal systems of fish.

Seasonal changes in reproductive-related vocal behavior are widespread among fishes. This review highlights recent studies of the vocal plainfin midshipman fish, Porichthys notatus, a neuroethological model system used for the past two decades to explore neural and endocrine mechanisms of vocal-acoustic social behaviors shared with tetrapods. Integrative approaches combining behavior, neurophysiology, neuropharmacology, neuroanatomy, and gene expression methodologies have taken advantage of simple, stereotyped and easily quantifiable behaviors controlled by discrete neural networks in this model system to enable discoveries such as the first demonstration of adaptive seasonal plasticity in the auditory periphery of a vertebrate as well as rapid steroid and neuropeptide effects on vocal physiology and behavior. This simple model system has now revealed cellular and molecular mechanisms underlying seasonal and steroid-driven auditory and vocal plasticity in the vertebrate brain.

Distinction of Fire Source from Smoke Using Discrete Probability Distribution and Neural Networks

Fire source distinction is useful in reducing the occurrence of false alarms and in choosing an effective fire extinguishing medium. This study was aimed at distinguishing fire sources through a smoke analysis. For this purpose, a smoke detection chamber was created, which was equipped with one light source and several light sensors for enabling simultaneous detection of light extinction and scattering, respectively. The test fires considered in this study had two kinds of sources: single fire source (paper, wood, and flammable liquid) and mixed fire source (paper-wood, paper-flammable liquid, and wood-flammable liquid mixtures). The amounts of extinction and scattering for each fire were measured experimentally, and the discrete probability distributions were calculated from the measured scattering amount. The optical characteristics of each fire were obtained using extinction data and the calculated probability distributions. These optical characteristics were then used for learning of neural networks, and the learned neural networks were used to distinguish fire sources from smoke generated in the case of both single fires and mixed fires. Results revealed that the neural networks could precisely distinguish fire sources on the basis of the smoke particles in the case of both single fires and mixed fires. The results of this study are expected to be useful in developing an advanced smoke detector that can distinguish fire sources in addition to detecting smoke.

Finite-time boundedness for uncertain discrete neural networks with time-delays and Markovian jumps

This paper is concerned with stochastic finite-time boundedness analysis for a class of uncertain discrete-time neural networks with Markovian jump parameters and time-delays. The concepts of stochastic finite-time stability and stochastic finite-time boundedness are first given for neural networks. Then, applying the Lyapunov approach and the linear matrix inequality technique, sufficient criteria on stochastic finite-time boundedness are provided for the class of nominal or uncertain discrete-time neural networks with Markovian jump parameters and time-delays. It is shown that the derived conditions are characterized in terms of the solution to these linear matrix inequalities. Finally, numerical examples are included to illustrate the validity of the presented results.

The Combined Forecasting Model of Discrete Verhulst-BP Neural Network Based on Linear Time-Varying

Firstly, this paper, aiming at the problem of errors produced by the transformation of differential equation directly into difference equation from traditional gray Verhulst model,through generating reciprocal for the original data sequence, constructs the discrete Vrhulst model based on linear time-varying（LTDVM model);And then we, taking the LTDVM predicted value as an input value and the original data as a mentor training value, put forward the combined forecasting model of discrete Verhulst-BP neural network based on linear time-varying. Meanwhile, in order to improve the training speed and agility and effectively avoid the saturation region of S-type function, this article normalized in advance the input data and mentor training values to better ensure the usefulness, self-learning ability and fault tolerance of the model. At last, we will study the cases to demonstrate that the model has high modeling and forecasting accuracy. DOI : http://dx.doi.org/10.11591/telkomnika.v12i4.4938

Convergence Theory for Unified Discrete-Time RNNs with Quasi-Symmetric Connection

This paper presents the global convergence theory of the discrete-time uniform pseudo projection anti-monotone network with the quasi–symmetric matrix, which removes the connection matrix constraints. The theory widens the range of applications of the discrete–time uniform pseudo projection anti–monotone network and is valid for many kinds of discrete recurrent neural network models.

Complex dynamics of a delayed discrete neural network of two nonidentical neurons

In this paper, we discover that a delayed discrete Hopfield neural network of two nonidentical neurons with self-connections and no self-connections can demonstrate chaotic behaviors. To this end, we first transform the model, by a novel way, into an equivalent system which has some interesting properties. Then, we identify the chaotic invariant set for this system and show that the dynamics of this system within this set is topologically conjugate to the dynamics of the full shift map with two symbols. This confirms chaos in the sense of Devaney. Our main results generalize the relevant results of Huang and Zou [J. Nonlinear Sci. 15, 291-303 (2005)], Kaslik and Balint [J. Nonlinear Sci. 18, 415-432 (2008)] and Chen et al. [Sci. China Math. 56(9), 1869-1878 (2013)]. We also give some numeric simulations to verify our theoretical results.

Global dynamics of discrete neural networks allowing non-monotonic activation functions

We show that in discrete models of Hopfield type some properties of global stability and chaotic behaviour are coded in the dynamics of a related one-dimensional equation. Using this fact, we obtain some new results on stability and chaos for a system of delayed neural networks; some relevant properties of our results are that we do not require monotonicity properties in the activation function, we allow any architecture of the network, and the conclusions are independent of the size of the time delays.

The paraventricular nucleus and heart failure

What is the topic of this review? This review gives an update on the cellular and molecular mechanisms within the autonomic nervous system involved in non-pathological and pathological cardiovascular regulation. What advances does it highlight? For cardiovascular homeostasis in non-pathological conditions to be maintained, discrete neural networks using specified signalling mechanisms at both cellular and molecular levels are required. In heart failure, the cell signalling protein partners CAPON and PIN decrease the bioavailability of nitric oxide by inhibiting neuronal nitric oxide synthase activity, leading to the removal of tonic neuronal inhibition. Following a myocardial infarction, pro-inflammatory cytokines in the paraventricular nucleus and the subsequent generation of reactive oxygen species, via angiotensin II activation of the angiotensin II type 1 receptor, increase neuronal excitability further, leading to sympathetic excitation. A pathological feature of heart failure is abnormal control of the sympathetic nervous system. The paraventricular nucleus of the hypothalamus (PVN) is one of the most important central sites involved in regulating sympathetic tone and is, in part, responsible for the dysregulation of the sympathetic nervous system evident in heart failure. Generation of sympathetic tone in response to fluctuations in cardiovascular regulation uses discrete anatomical pathways and neurochemical modulators. Direct and indirect projections from the PVN pre-autonomic neurons innervate the sympathetic preganglionic neurons in the spinal cord, which in turn innervate sympathetic ganglia that give rise to the sympathetic nerves. Pre-autonomic neurons of the PVN themselves receive an afferent input arising from the nucleus tractus solitarii, and viscerosensory receptors convey cardiovascular fluctuations to the nucleus tractus solitarii. The PVN contains excitatory and inhibitory neurons, whose balance determines the sympathetic tone. In non-pathological conditions, the tonic inhibition of the PVN pre-autonomic neurons is mediated by GABA- and NO-releasing neurons. In heart failure, the pre-autonomic neurons are disinhibited by the actions of the excitatory neurotransmitters glutamate and angiotensin II, leading to increased sympathetic activity. A key feature of the disinhibition is a reduction in the bioavailability of NO as a consequence of disrupted CAPON and PIN signalling mechanisms within the neuron. Another critical feature that contributes to increased neuronal excitation within the PVN is the production of pro-inflammatory cytokines immediately following a myocardial infarction, the activation of the angiotensin II type 1 receptor and the production of reactive oxygen species. By examining the changes associated with the sympathetic nervous system pathway, we will progress our understanding of sympathetic regulation in heart failure, identify gaps in our knowledge and suggest new therapeutic strategies.

Discrete Process Neural Networks Algorithm with Application

To address the training of discrete process neural networks, two algorithms based on numerical integral are proposed. The cubic spline integral and the parabolic interpolation integral are directly applied to dealing with the aggregation of discrete samples and weights in time-domain, and the Levenberg-Marquard algorithm is employed to adjusting the networks parameters. The effectiveness of the proposed algorithms is testified by applying the networks to the caravan insurance policy prediction, the experimental results show that the performance of two algorithms is relatively close, and superior to the orthogonal basis expansion-based process neural networks.

НЕЙРОННЫЕ СЕТИ ХЕММИНГА И ХЕББА, СПОСОБНЫЕ ДООБУЧАТЬСЯ

Or the first time architectures and algorithms of the discrete stable and plastic neural networks based on neural networks Hemming and Hebb are proposed. And these algorithms and architectures can not only to learn, but also to recognize new information. New networks can be an alternative to discrete neural network adaptive resonance theory. Figs.: 5. Tabl.: 4. Refs.: 10 titles.

Individual Differences in Attentional Bias Associated with Cocaine Dependence Are Related to Varying Engagement of Neural Processing Networks

Cocaine and other drug dependencies are associated with significant attentional bias for drug use stimuli that represents a candidate cognitive marker of drug dependence and treatment outcomes. We explored, using fMRI, the role of discrete neural processing networks in the representation of individual differences in the drug attentional bias effect associated with cocaine dependence (AB-coc) using a word counting Stroop task with personalized cocaine use stimuli (cocStroop). The cocStroop behavioral and neural responses were further compared with those associated with a negative emotional word Stroop task (eStroop) and a neutral word counting Stroop task (cStroop). Brain-behavior correlations were explored using both network-level correlation analysis following independent component analysis (ICA) and voxel-level, brain-wide univariate correlation analysis. Variation in the attentional bias effect for cocaine use stimuli among cocaine-dependent men and women was related to the recruitment of two separate neural processing networks related to stimulus attention and salience attribution (inferior frontal-parietal-ventral insula), and the processing of the negative affective properties of cocaine stimuli (frontal-temporal-cingulate). Recruitment of a sensory-motor-dorsal insula network was negatively correlated with AB-coc and suggested a regulatory role related to the sensorimotor processing of cocaine stimuli. The attentional bias effect for cocaine stimuli and for negative affective word stimuli were significantly correlated across individuals, and both were correlated with the activity of the frontal-temporal-cingulate network. Functional connectivity for a single prefrontal-striatal-occipital network correlated with variation in general cognitive control (cStroop) that was unrelated to behavioral or neural network correlates of cocStroop- or eStroop-related attentional bias. A brain-wide mass univariate analysis demonstrated the significant correlation of individual attentional bias effect for cocaine stimuli with distributed activations in the frontal, occipitotemporal, parietal, cingulate, and premotor cortex. These findings support the involvement of multiple processes and brain networks in mediating individual differences in risk for relapse associated with drug dependence.