Learning Vector Quantization Network Research Articles

UCAV situation assessment method based on C-LSHADE-Means and SAE-LVQ

The unmanned combat aerial vehicle (UCAV) is a research hot issue in the world, and the situation assessment is an important part of it. To overcome shortcomings of the existing situation assessment methods, such as low accuracy and strong dependence on prior knowledge, a data-driven situation assessment method is proposed. The clustering and classification are combined, the former is used to mine situational knowledge, and the latter is used to realize rapid assessment. Angle evaluation factor and distance evaluation factor are proposed to transform multi-dimensional air combat information into two-dimensional features. A convolution success-history based adaptive differential evolution with linear population size reduction-means (C-LSHADE-Means) algorithm is proposed. The convolutional pooling layer is used to compress the size of data and preserve the distribution characteristics. The LSHADE algorithm is used to initialize the center of the mean clustering, which overcomes the defect of initialization sensitivity. Comparing experiment with the seven clustering algorithms is done on the UCI data set, through four clustering indexes, and it proves that the method proposed in this paper has better clustering performance. A situation assessment model based on stacked autoencoder and learning vector quantization (SAE-LVQ) network is constructed, and it uses SAE to reconstruct air combat data features, and uses the self-competition layer of the LVQ to achieve efficient classification. Compared with the five kinds of assessments models, the SAE-LVQ model has the highest accuracy. Finally, three kinds of confrontation processes from air combat maneuvering instrumentation (ACMI) are selected, and the model in this paper is used for situation assessment. The assessment results are in line with the actual situation.

Variants of recurrent learning vector quantization

Learning Vector Quantization (LVQ) and its cost-function-based variant called Generalized Learning Vector Quantization (GLVQ) are powerful, yet simple and interpretable Machine Learning (ML) algorithms for multi-class classification. Even though GLVQ is an effective tool for classifying vectorial data in its native form, it is not particularly suited to handle raw sequence data of potentially different lengths. Usually, this problem is addressed by manually engineering fixed-length features from the raw data or by employing recurrent networks. Therefore, a natural idea is to try to incorporate recurrent units for data processing into the GLVQ network structure. The processed data can then be compared in a latent space for classification decisions. Yet, not much work has been done in direction as far as we are aware. Existing methods to handle sequential data in the framework of GLVQ are rather unsophisticated and severely outdated. In this paper, we provide a general framework for incorporating recurrent structures in an LVQ network and derive two classification models as variants of Recurrent Learning Vector Quantization, namely RecLVQ and LVQRNN. We also demonstrate the abilities of such approaches on illustrative classification problems.

Information System to Analyze Human Environmental Wellbeing for Assessing and Reducing Technogenic Risk

The article provides a rationale for the need of developing the information system to analyze human environmental wellbeing (HEW) on the basis of artificial neural networks. The application of artificial neural networks is pre-conditioned by the need of prompt analysis and processing to take decisions on the implementation of environment protection measures for large data arrays accumulated as a result of environmental monitoring. The suggested information system for technogenic risk assessment implements a multi-limit way to control the state of a multi-parametric object and allows assessing human environmental wellbeing in real time. The authors provide structural charts for the information system to assess technogenic risk and the block to process and store information on the basis of an artificial neural network LVQ (Learning Vector Quantization). The authors provide an algorithm and the results for the artificial neural network LVQ training.

Energy consumption, flow characteristics and energy-efficient design of cup-shape blade stirred tank reactors: Computational fluid dynamics and artificial neural network investigation

Developing high-efficient and green energy-saving stirred tank reactors is of vital importance in liquid-liquid processing industries. Therefore, this work revealed the energy consumption, flow characteristics, and explored the energy-saving strategies of novel-designed Cup-shape Blade (CB) stirred tank reactors characterized by easy fabrication, low energy consumption and large pumping flow rate. Computational Fluid Dynamics (CFD) and Artificial Neural Network (ANN) were conducted to investigate the influences of geometrical parameters on liquid-liquid power number (NP), flow number (NQ) and flow patterns. The improved method for hyper-parameter specification indicates that the Bayesian regulation training algorithm and the two-hidden-layer scheme possess the best performance. Back Propagation Neural Network (BPNN) presents the best predictive accuracy for NP and NQ, and Learning Vector Quantization (LVQ) network shows the best classification capability of flow patterns. Analyses based on the ANN modeling and the “Local Interpretable Model-agnostic Explanations” method show the joint and complicated effects of geometrical parameters on energy consumption and flow characteristics. Procedures were proposed to ascertain the energy-efficiency blade configuration based on ANN.

Validation of Neural Network Predictions for the Outcome of Refractive Surgery for Myopia

Background: Refractive surgery (RS) for myopia has made a very big progress regarding its safety and predictability of the outcome. Still, a small percentage of operations require retreatment. Therefore, both legally and ethically, patients should be informed that sometimes a corrective RS may be required. We addressed this issue using Neural Networks (NN) in RS for myopia. This was a recently developed validation study of a NN. 
 Methods: We anonymously searched the Ophthalmica Institute of Ophthalmology and Microsurgery database for patients who underwent RS with PRK, LASEK, Epi-LASIK or LASIK between 2010 and 2018. We used a total of 13 factors related to RS. Data was divided into four sets of successful RS outcomes used for training the NN, successful RS outcomes used for testing the NN performance, RS outcomes that required retreatment used for training the NN and RS outcomes that required retreatment used for testing the NN performance. We created eight independent Learning Vector Quantization (LVQ) networks, each one responding to a specific query with 0 (for the retreat class) or 1 (for the correct class). The results of the 8 LVQs were then averaged so we could obtain a best estimate of the NN performance. Finally, a voting procedure was used to reach to a conclusion.
 Results: There was a statistically significant agreement (Cohen’s Kappa = 0.7658) between the predicted and the actual results regarding the need for retreatment. Our predictions had good sensitivity (0.8836) and specificity (0.9186).
 Conclusion: We validated our previously published results and confirmed our expectations for the NN we developed. Our results allow us to be optimistic about the future of NNs in predicting the outcome and, eventually, in planning RS.

Variants of DropConnect in Learning vector quantization networks for evaluation of classification stability

Dropout and DropConnect are useful methods to prevent multilayer neural networks from overfitting. In addition, it turns out that these tools can also be used to estimate the stability of networks regarding disturbances. Prototype based networks gain more and more attraction in current research because of their inherent interpretability and robust behavior. Popular prototype-based classifiers are support vector machines and the heuristically motivated Learning Vector Quantizer (LVQ). The Generalized Matrix LVQ (GMLVQ) is an extension of LVQ which can be interpreted as a special multilayer network containing a projection and a prototype layer. First in this paper, we extend the linear projection layer of GMLVQ to a non-linear mapping by employing different non-linear activations functions. Second, we compare the classification decision stabilities of the linear and the non-linear GMLVQ regarding DropConnect while taking the neural network perspective. Thus we can adopt DropConnect ideas known from multilayer perceptron learning to investigate stability and robustness of GMLVQ. To this end, the evaluation of the stability is done in terms of a information theoretic stability measure based on the Shannon-Entropy. We demonstrate the approach for three real world data sets from Raman spectroscopy, multi-spectral remote sensing and the well-known MNIST data set.

PSO–SOM Neural Network Algorithm for Series Arc Fault Detection

Self-organizing feature map (SOM) neural network is a kind of competitive neural network with unsupervised learning. It has the strong abilities of self-organization and self-learning. However, the classification accuracy of SOM neural network may decrease when the features of tested object are not obvious. In this paper, the particle swarm optimization (PSO) algorithm is used to optimize the weight values of SOM network. Three indexes, i.e., intra-class density, standard deviation and sample difference, are used to judge the weight value, which can improve the classification accuracy of the SOM network. PSO–SOM network is applied to the detection of series arc fault in electrical circuits and compared with conventional SOM network and learning vector quantization (LVQ) network. The detection accuracy of the PSO–SOM network is 95%, which is higher than conventional SOM network and LVQ network.

Predictive modelling of injury severity in bicycle–motor vehicle collisions utilizing learning vector quantization: a case study of Britain’s cycling capital

Cambridge is truly known as Britain’s cycling capital but this has all been achieved without any real cycling infrastructure. Therefore, safety concern is the persistent barrier to cycling in Cambridge and its reputation is on the decline. Thus in response to the small number of the literatures on prediction models for cyclist related injuries, this study presents a modelling technique that applies learning vector quantization network to predict injury severity of cyclist. By discovering the potentially relationships between the injury levels and the factors that contribute to their generation, the model predicts the likelihood of the injury into two classes: slight and killed/seriously injured. The findings display that the effect of junction actions are almost double. Following this, T/staggered junction on an unclassified bend was discovered as UK's cycling capital’s “collision hotspots”. Subsequently, absence of a sufficient number of crossing facilities for cyclists had the largest effect. All other things being held near equal; rush hours during the weekdays, vehicle manoeuvre due to a poor turn and parked vehicle, going ahead bend manoeuvre by cyclist, limited modern protected bike lanes, wet road surface owing to inclement weather, vehicle blind spot and driving too close, visual distraction caused by adverse lighting condition and dazzle of sun, and junction control commonly at situations of give-way or uncontrolled intersections. The study then ends by maximising the overall predictive accuracy through contacting the most sensitive predictors into the model. Consequently, a professional road safety education needs to be delivered so as to crack down the human failure as a main contributory factor.

Application of Extension Neural Network with Discrete Wavelet Transform and Parseval’s Theorem for Power Quality Analysis

The development of renewable energy and the increase of intermittent fluctuating loads have affected the power quality of power systems, and in the long run, damage the power equipment. In order to effectively analyze the quality of power signals, this paper proposes a method of signal feature capture and fault identification, as based on the extension neural network (ENN) algorithm combined with discrete wavelet transform (DWT) and Parseval’s theorem. First, the original power quality disturbance (PQD) transient signal was subjected to DWT, and its spectrum energy was calculated for each order of wavelet coefficients through Parseval’s theorem, in order to effectively intercept the eigenvalues of the original signal. Based on the features, the extension neural algorithm was used to establish a matter-element model of power quality disturbance identification. In addition, the correlation degree between the identification data and disturbance types was calculated to accurately identify the types of power failure. To verify the accuracy of the proposed method, five common power quality disturbances were analyzed, including voltage sag, voltage swell, power interruption, voltage flicker, and power harmonics. The results were then compared with those obtained from the back-propagation network (BPN), probabilistic neural network (PNN), extension method and a learning vector quantization network (LVQ). The results showed that the proposed method has shorter computation time (0.06 s), as well as higher identification accuracy at 99.62%, which is higher than the accuracy rates of the other four types.

Neural network classification of surface quality after hard turning of 105WCr6 steel

The paper presents the results of a surface quality study after hard turning on a CNC lathe. Ring workpieces made of 105WCr6 steel and hardened to HRC 55 are used in this work. Data was obtained on surface quality and type of chips in a three-factor experiment for end face cutting. In order to assess the surface quality, it was photographed on an optical microscope with 4, 10, 40 times magnification. The surface quality was evaluated by traces of processing and divided into three types: the absence of moire, a clear moire, and an intermediate type of surface. The chip morphology was divided into the following categories: discontinuous, snarled and ribbon chips. To predict both parameters for different cutting conditions artificial neural networks (ANNs) were used. Different ANNs are applied to achieve the best classification results. In this work probabilistic neural network (PNN), feedforward network and learning vector quantization (LVQ) network are used. The results of modeling all networks are similar and can be used for technological preparation of production.

Using neural networks to predict the outcome of refractive surgery for myopia

Introduction: Refractive Surgery (RS), has advanced immensely in the last decades, utilizing methods and techniques that fulfill stringent criteria for safety, efficacy, cost-effectiveness, and predictability of the refractive outcome. Still, a non-negligible percentage of RS require corrective retreatment. In addition, surgeons should be able to advise their patients, beforehand, as to the probability that corrective RS will be necessary. The present article addresses these issues with regard to myopia and explores the use of Neural Networks as a solution to the problem of the prediction of the RS outcome. Methods: We used a computerized query to select patients who underwent RS with any of the available surgical techniques (PRK, LASEK, Epi-LASIK, LASIK) between January 2010 and July 2017 and we investigated 13 factors which are related to RS. The data were normalized by forcing the weights used in the forward and backward propagations to be binary; each integer was represented by a 12-bit serial code, so that following this preprocessing stage, the vector of the data values of all 13 parameters was encoded in a binary vector of 1 × (13 × 12) = 1 × 156 size. Following the preprocessing stage, eight independent Learning Vector Quantization (LVQ) networks were created in random way using the function Ivqnet of Matlab, each one of them responding to one query with (0 retreat class) or (1 correct class). The results of the eight LVQs were then averaged to permit a best estimate of the network’s performance while a voting procedure by the neural nets was used to arrive at the outcome Results: Our algorithm was able to predict in a statistically significant way (as evidenced by Cohen’s Kappa test result of 0.7595) the need for retreatment after initial RS with good sensitivity (0.8756) and specificity (0.9286). Conclusion: The results permit us to be optimistic about the future of using neural networks for the prediction of the outcome and, eventually, the planning of RS.

On the feasibility of metal oxide gas sensor based electronic nose software modification to characterize rice ageing during storage

The aroma change of aromatic and non-aromatic rice was traced during storage using electronic nose (E-nose). Various steady-state and transient features were derived from the adsorption and desorption phases of the E-nose responses to characterize the ageing process and classify the storage durations. Principal component analysis was utilized to analyse the ageing in terms of seven classes related to the storage duration of 6 months. The aromatic samples followed a specific pathway in its timeline with separately distinctive grouping, revealing the reduction of ageing indices. The aromatic rice underwent crucial changes in its volatile compounds, mainly in the early stages of storage. While, confused grouping of the non-aromatic rice proved its stability due to less diverse range of the aroma compounds. Several artificial neural networks, namely back propagation (BP), radial basis function (RBF), and learning vector quantization (LVQ) were used to classify the storage durations. For the aromatic samples, full classification was achieved by using all the networks. For the non-aromatic rice, the developed RBF and LVQ networks represented full classification by using the features of rise time and polynomial curve fitting parameters. It is concluded that the E-nose system along with the developing methods could be utilized to control the rice ageing.

A LVQ-like Network Approach for Classification

The Learning Vector Quantization (LVQ) network is a hybrid network, employing both unsupervised and supervised learning to solve classification problems. It is simple, but some issues exist. For example, the number of hidden nodes has to be decided in advance, and the spread of the data is not considered. In this paper, we propose a LVQ-like network approach to remedy these disadvantages. An iterative self-constructing clustering algorithm is used to determine the number of hidden nodes in the hidden layer. Data are described by clusters with appropriate centers and deviations. As a result, the number of hidden nodes is determined automatically. Also, through the incorporation of adaptive deviations, data can be described more appropriately. Experimental results are shown to demonstrate the effectiveness of the proposed approach.

Study of Face Orientation Recognition Based on Neural Network

Learning vector quantization (LVQ) network and back-propagation (BP) network are constructed easily making use of MATLAB toolbox on the basis of maintaining the recognition rate. Face images are randomly selected from images set as training data of LVQ network and BP network. LVQ algorithm and BP algorithm are used to train network. The automatic recognition of face orientation is realized when the system obtains convergence network. First, all images are processed by edge detection. Then feature vectors representing position of the eye were extracted from edge detected images. Feature vectors of training set are sent to network to adjust the parameters which ensures the convergence speed and performance of the network. Experimental results show that the constructed LVQ network and BP network can judge face orientation according to feature vectors of input images. Generally, the recognition rate of LVQ network is higher than that of BP network. The LVQ network and BP network are both feasible and effective for face orientation recognition to some extent. The advantage of this work is that the recognition system is efficient and easy to promote. This paper focuses on how to use MATLAB easily to design identification network rather than the complexity of identification system. The future research will focus on the stability and robustness of recognition network.

Pulmonary nodules computer-aided diagnosis based on feature integration and ABC-LVQ network

For the computer aided diagnosis of lung cancer, a malignancy identification method based on multi-feature integration and learning vector quantisation (LVQ) network optimised by artificial bee colony (ABC) is proposed in this work. Firstly, the traditional features and the hidden features learned by Sparse Autoencoder of nodules are respectively extracted, and then the canonical correlation analysis (CCA) is used for feature integration. For classification, the ABC algorithm is used to optimise the LVQ network to overcome its sensitivity to initial value. Finally, the integrated features of nodules are input into the optimised classifier and the diagnosis results are obtained. Experimental results on LIDC pulmonary nodule image datasets show that this method can effectively identify the malignancy of nodules, with the area under the receiver operating characteristic (ROC) curve (AUC) of 0.90, 0.83, 0.80, 0.80, 0.85 for nodules of malignancy 1-5 classification, respectively.

Pulmonary nodules computer-aided diagnosis based on feature integration and ABC-LVQ network

For the computer aided diagnosis of lung cancer, a malignancy identification method based on multi-feature integration and learning vector quantisation (LVQ) network optimised by artificial bee colony (ABC) is proposed in this work. Firstly, the traditional features and the hidden features learned by Sparse Autoencoder of nodules are respectively extracted, and then the canonical correlation analysis (CCA) is used for feature integration. For classification, the ABC algorithm is used to optimise the LVQ network to overcome its sensitivity to initial value. Finally, the integrated features of nodules are input into the optimised classifier and the diagnosis results are obtained. Experimental results on LIDC pulmonary nodule image datasets show that this method can effectively identify the malignancy of nodules, with the area under the receiver operating characteristic (ROC) curve (AUC) of 0.90, 0.83, 0.80, 0.80, 0.85 for nodules of malignancy 1-5 classification, respectively.

Comparative study of LVQ and BPN ECG classifier

ECG is the electrical waveform of heart activity. It contains much information on heart disease. It is very important to diagnosis the heart disease as soon as possible otherwise it can be harmful to patient. This paper presents to classify ECG signal using learning vector quantisation and back propagation neural network and feature of ECG (morphology and frequency domain) features. In this paper, the 45 ECG signals from MIT-BIH arrhythmia database are used to classify in to two classes, one is normal and another one is abnormal using above mentioned classifier. Out of 45 signals 25 are normal and 20 are abnormal according to MIT-BIH. Twenty-eight morphological features and four frequency domain features are set as an input to the classifier. The performance of classifier measures in the terms of sensitivity (SE), positive predictivity (PP) and specificity (SP). The system performance is achieved with 82.35% accuracy using LVQ and 94.11% using BPN.

A novel hybrid strategy for PM2.5 concentration analysis and prediction

The analysis and prediction of air pollutants are of great significance in environmental research today since airborne pollution is a substantial threat, especially in urban agglomerations of China. To develop more effective warning systems and management advice, the authorities and city dwellers need more accurate forecasts of the air pollution. Most previous analysis systems were based on costly observation apparatus at fixed sites, forecasting models were usually built on observations within a certain range, and some observations contained biases. In this paper, a novel and effective framework, termed HML-AFNN, was successfully developed to analyse and forecast the concentration of particular matter (PM2.5) for a selected number of forward time steps. In a simulation of the trajectory of air pollutants, the high-dimension association rules (HDAR) approach considered the tempo-spatial relations, as well as the meteorological and geographical factors of the ambient regions, as parameters. In addition, the learning vector quantization (LVQ) network was adopted to select the appropriate inputs to improve the efficiency of the training process. Moreover, an adaptive fuzzy neural network (AFNN), a combination of neural and fuzzy logic, was utilized to analyse and predict the PM2.5 concentration. The experiment results of our study on two major urban agglomerations of China, the Jing-Jin-Ji area and Pearl River Delta, over a period of more than one year demonstrated that the developed hybrid HML-AFNN model outperforms a plain AFNN, an HM-AFNN model without LVQ and the least squares support vector machines (LS-SVM); this superior performance can be determined from the values of several error indexes, including MAE, MAPE and band errors. This hybrid model, which has robust and accurate results, shows the potential to be a political and administrative method to issue effective early warnings and to design suitable abatement strategies.

A Synthetic Player for Ayὸ Board Game Using Alpha-Beta Search and Learning Vector Quantization

Game playing especially, Ayὸ game has been an important topic of research in artificial intelligence and several machine learning approaches have been used, but the need to optimize computing resources is important to encourage significant interest of users. This study presents a synthetic player (Ayὸ) implemented using Alpha-beta search and Learning Vector Quantization network. The program for the board game was written in Java and MATLAB. Evaluation of the synthetic player was carried out in terms of the win percentage and game length. The synthetic player had a better efficiency compared to the traditional Alpha-beta search algorithm.

Intrinsic Plasticity for Natural Competition in Koniocortex-Like Neural Networks.

In this paper, we use the neural property known as intrinsic plasticity to develop neural network models that resemble the koniocortex, the fourth layer of sensory cortices. These models evolved from a very basic two-layered neural network to a complex associative koniocortex network. In the initial network, intrinsic and synaptic plasticity govern the shifting of the activation function, and the modification of synaptic weights, respectively. In this first version, competition is forced, so that the most activated neuron is arbitrarily set to one and the others to zero, while in the second, competition occurs naturally due to inhibition between second layer neurons. In the third version of the network, whose architecture is similar to the koniocortex, competition also occurs naturally owing to the interplay between inhibitory interneurons and synaptic and intrinsic plasticity. A more complex associative neural network was developed based on this basic koniocortex-like neural network, capable of dealing with incomplete patterns and ideally suited to operating similarly to a learning vector quantization network. We also discuss the biological plausibility of the networks and their role in a more complex thalamocortical model.