Competitive Layer Research Articles

Моделювання процесу ранжування у нейромережному класифікаторі об’єктів

As part of expert systems for various purposes, one of the basic ones is the decision support subsystem, which, in turn, requires the need for a procedure for classifying ob-jects. This is especially evident in intelligent medical diagnostic systems, which widely use artificial intelligence methods and tools. In this context, an approach involving mod-ern neurotechnology methods has proven to be effective at a high level. This paper con-siders a variant of the structural organization of a neural network classifier of objects as an improved model of the Hamming neural network. The peculiarity of this variant of the classifier is the expansion of its functionality by forming the ranks of the classified ob-ject in all defined classes. In the case of medical diagnosis, this means ranking all possi-ble diagnoses of a disease, i.e. determining not only the most likely diagnosis, but also the closest in rank to it. In fact, this will allow us to clarify the diagnosis, and thus im-prove the results of medical diagnosis. Accordingly, we simulated the classification pro-cess with the ranking of results, which corresponds to the classification with the realiza-tion of competition between the neurons of the competitive layer using negative-reverse (lateral) connections. This approach is basic in the theory of neural networks for deter-mining the winning neuron according to the WTA (Winner Takes All). Simulation model-ing of the classification variant was performed using specific biomedical data (eight symptoms) for the diagnosis of appendicitis (four diagnoses). The results of modeling the processes of neural network classification of objects with the formation of appropri-ate ranks are presented in the form of a table. They confirmed the correctness of the functioning algorithm for the considered classification model.

Boosting the Development and Management of Wind Energy: Self-Organizing Map Neural Networks for Clustering Wind Power Outputs

Aimed at the information loss problem of using discrete indicators in wind power output characteristics analysis, a self-organizing map neural network-based clustering method is proposed in this study. By identifying the appropriate representativeness and topological structure of the competition layer, cluster analysis of the wind power output process in four seasons is realized. The output characteristics are evaluated through multiple evaluation indicators. Taking the wind power output of the Hunan power grid as a case study, the results underscore that the 1 × 3-dimensional competition layer structure had the highest representativeness (72.9%), and the wind power output processes of each season were divided into three categories, with a robust and stable topology structure. Summer and winter were the most representative seasons. Summer had strong volatility and small wind power outputs, which required the utilization of other power sources to balance power supply and load demand. Winter featured low volatility and large wind power outputs, necessitating cooperation with peak-shaving power sources to enhance the power grid’s absorbability to wind power. The seasonal clustering analysis of wind power outputs will be helpful to analyze the seasonality of wind power outputs and can provide scientific and technical support for guiding the power grid’s operation and management.

Tailoring ZrWNb-Based Refractory High Entropy Alloys: A Multidirectional Characterization Through Elemental, Physical, Thermodynamic, and Nuclear Radiation Attenuation Properties

This study explores the development and characterization of 18 novel ZrWNb[(X)(Y)] (where XY are Hf, Ta, Mo, V, Ti) refractory high entropy alloys (RHEAs) designed to enhance mechanical robustness, thermodynamic stability, and radiation shielding effectiveness. Through a rigorous analysis, we evaluated the elastic modulus, entropy, and enthalpy of mixing, atomic size difference, valence electron concentration, and the omega parameter to understand the alloys’ phase behavior and structural integrity. Our findings revealed a broad range of elastic modulus values, indicating diverse mechanical adaptability suitable for high-stress applications. The thermodynamic assessment highlighted several RHEAs with favorable phase stability, particularly ZrWNbTaHf and ZrWNbTiTaHf, which are poised for high-performance usage due to their balanced mixing entropy and enthalpy. This study also benchmarks the RHEAs against conventional alloys, with sample R1 exhibiting the lowest fast neutron effective removal cross section, underscoring its superior neutron shielding capabilities. Additionally, R1 demonstrated exceptional photon attenuation, as evidenced by its competitive half-value layer performance across an extensive energy spectrum. Collectively, these results position R1 as a standout candidate, offering a significant advancement in materials science for applications demanding stringent radiation shielding, such as in nuclear reactor environments and space exploration.

Nitrogen doped single layer graphene for CZTS-based thin film solar cells

CdS thin films are commonly utilized as a buffer layer in chalcopyrite thin-film solar cells. However, due to the toxic nature of Cadmium (Cd), ongoing efforts are being directed towards exploring alternative options. In this contribution, doped graphene film with remarkable optical and electrical properties has been introduced for the first time as an alternative buffer layer, replacing CdS in CZTS thin-film solar cell application. In this study, nitrogen-doped graphene (N-doped graphene) film was utilized as a substitute buffer layer in the CZTS thin-film solar cell structure, replacing the conventional CdS thin film. For comparative analysis, CZTS/N-doped graphene and CZTS/CdS traditional solar cell structures were fabricated and separately characterized. The CZTS thin films produced were examined through EDX, XRD, SEM, Raman, optical transmission and PL spectroscopy measurements. According to performed analyses, the Cu-poor and Zn-rich kesterite CZTS thin films exhibited a uniform and dense polycrystalline microstructure as observed in surface and cross-sectional SEM images. XRD spectra of the kesterite CZTS thin film displayed predominant peaks corresponding to the (112), (220/204), and (312/116) diffraction planes of the kesterite CZTS phase. Raman spectra showed a dominant peak at ∼336 cm−1 associated with the kesterite CZTS phase. PL emission spectra indicated transitions from the conduction band to defect levels. The CVD-grown doped graphene film exhibited a 3.43 I2D/IG ratio and a 25 cm−1 FWHM of the 2D peak, indicating a single-layer graphene according to Raman analysis. Permanent nitrogen doping with a 2% atomic concentration was confirmed by XPS measurement. The optical transmission measurement of the single layer doped graphene film showed a 95% transmittance value. Nitrogen doping was contributed to decrease the sheet resistance of the graphene film. The Glass/Mo/CZTS/N-doped graphene/i-ZnO/ITO/Al solar cell displayed a VOC of 0.267 V, JSC of 24.6 mA/cm2, FF of 36.46, and η of 2.37%, showing higher FF and Jsc values but lower conversion efficiency compared to the Glass/Mo/CZTS/CdS/i-ZnO/ITO/Al traditional solar cell structure. Hence, the superior working function and transparency properties position the N-doped graphene film as a competitive buffer layer for use in CZTS-based thin-film solar cells.

Gradient-Based Competitive Learning: Theory

Deep learning has been recently used to extract the relevant features for representing input data also in the unsupervised setting. However, state-of-the-art techniques focus mostly on algorithmic efficiency and accuracy rather than mimicking the input manifold. On the contrary, competitive learning is a powerful tool for replicating the input distribution topology. It is cognitive/biologically inspired as it is founded on Hebbian learning, a neuropsychological theory claiming that neurons can increase their specialization by competing for the right to respond to/represent a subset of the input data. This paper introduces a novel perspective by combining these two techniques: unsupervised gradient-based and competitive learning. The theory is based on the intuition that neural networks can learn topological structures by working directly on the transpose of the input matrix. At this purpose, the vanilla competitive layer and its dual are presented. The former is representative of a standard competitive layer for deep clustering, while the latter is trained on the transposed matrix. The equivalence of the layers is extensively proven both theoretically and experimentally. The dual competitive layer has better properties. Unlike the vanilla layer, it directly outputs the prototypes of the data inputs, while still allowing learning by backpropagation. More importantly, this paper proves theoretically that the dual layer is better suited for handling high-dimensional data (e.g., for biological applications), because the estimation of the weights is driven by a constraining subspace which does not depend on the input dimensionality, but only on the dataset cardinality. This paper has introduced a novel approach for unsupervised gradient-based competitive learning. This approach is very promising both in the case of small datasets of high-dimensional data and for better exploiting the advantages of a deep architecture: the dual layer perfectly integrates with the deep layers. A theoretical justification is also given by using the analysis of the gradient flow for both vanilla and dual layers.

A novel multi-scale competitive network for fault diagnosis in rotating machinery

Bearing fault diagnosis plays a vital role in ensuring the safe and reliable operation of rotating machinery. The diagnostic process is more difficult when the fault is in its early stages, as fewer fault components are contained in the vibration signal, making the diagnosis process more difficult. To improve the accuracy and efficiency of fault type identification, a novel multi-scale competitive network for fault diagnosis is proposed in this paper. First, to obtain multi-scale features and fully utilize the features in the intermediate layer, ensuring the completeness of fault information, a novel improved multi-scale feature fusion residual network (IMSFFRN) is proposed to exploit deep features for vibration signals. Specifically, multi-scale features are obtained by convolution with different dilation rates, and features from adjacent intermediate layers are selected for efficient fusion. Second, different features have varying importance in fault detection tasks. To make neurons more sensitive to specific faults, we propose a multiple-winning consciousness self-organizing map (MCSOM) competition layer, in which each neuron learns specific faults through competition, and the neuron that wins the competition updates its weights. This distinguishes the sensitivity of neurons to different faults. Finally, the generalizability of the network is improved by using support vector machines (SVMs) to classify fault classes. To affirm the efficacy of the proposed approach, a comprehensive evaluation is conducted on the CWRU, PU bearing dataset and SEU gear dataset. The results indicate that the accuracies achieved by the method proposed in this paper are 100%, 99.56% and 100%, respectively. It is superior to other methods proposed in this paper. Furthermore, it is observed that the proposed method exhibits high robustness in noisy environments.

Linear quadratic mean-field game-team analysis: A mixed coalition approach

Mean-field theory has been extensively explored in decision analysis of large-scale (LS) systems but traditionally in “pure” cooperative or competitive settings. This leads to the so-called mean-field game (MG) or mean-field team (MT). This paper introduces a new class of LS systems with cooperative inner layer and competitive outer layer, so a “mixed” mean-field analysis is proposed for distributed game-team strategy. A novel asymptotic mixed-equilibrium-optima is also proposed and verified.

Measuring the Maturity Level of Oil Palm Fruit For CPO Production Based on Color With Using the LVQ Method

Palm oil is a very important commodity besides oil and gas which also has a fairly good export value. The palm oil produced must be supported by the quality standards set by SNI. The level of maturity when harvesting oil palm fruit greatly influences the quality of Crude Palm Oil (CPO) production, which is crude palm oil that has a reddish color obtained from extraction or from the pressing process of oil palm fruit flesh. In fact, in the field of oil palm fruit harvest, there are still many oil palm fruit that are not ripe enough and can even be said to be still raw, entering the CPO production process. Determination of the maturity level of oil palm fruit is generally determined based on the amount of loose fruit and color, so handling the harvest of oil palm fruit is an important activity in improving the quality of CPO. It is necessary to build a system capable of managing and processing palm fruit images to measure the maturity level of the palm fruit to be produced. To obtain the right level of accuracy, this research uses the Learning Vector Quantization (LVQ) method. LVQ is a method for conducting supervised competitive layer learning. From the results of trials conducted, it is proven that the system can measure very ripe oil palm fruit with HSV values (0.052209; 0.896021; 0.791114).

Effect of SO2 on HCl removal over ethanol-hydrated CaO adsorbent: Mechanism of competitive adsorption and product layer shielding

The elimination of hydrogen chloride (HCl) is essential for wastewater sequestration and safe operation of equipment in coal-fired power plants. A new adsorbent of ethanol-hydrated CaO has been developed, which possesses a 100% HCl removal efficiency lasting 30 min in a fixed bed reactor being much superior to traditional dechlorinators. Nevertheless, SO2 in flue gas has instinctive and strong competition with HCl on the alkaline adsorbent that results in detrimental effects and makes the removal mechanisms confused and unclear. To solve this issue, a series of experiments were designed to evaluate how SO2 influences the performance of HCl adsorption over the adsorbent. Comprehensive characterizations and density functional theory (DFT) calculations further studied the mechanisms of the competition mechanism between SO2 and HCl. The results indicate that SO2 has a detrimental impact on HCl removal by consuming the active Ca site, diminishing the alkalinity and worsening the surface activity due to the product layer formation. DFT calculations reveal that the high adsorption energy of SO2 suppresses the active sites for HCl removal. The faded surface electrostatic potentials (ESP) around the sulfurized Ca active sites lead to apparent activity shielding on the absorbent surface. Besides, the frontier molecular orbital (FMO) theory unveils a narrow electronic transfer energy gap for SO2 adsorption. These findings contribute to elucidation of SO2 competition to HCl on the calcium-based adsorbent in coal-fired flue gas.

Millimeter Channel Clustering by Self-Organizing Maps With Time-Varying Topological Structure

Cluster-based channel model plays an important role in the fifth-generation (5G) and beyond 5G (B5G) wireless communication systems. A well-designed clustering algorithm can achieve a good compromise between the accuracy and complexity of the wireless channel model. The dependence on prior information of cluster, such as the number of clusters, is a common dilemma of current algorithms. In this article, a self-organizing map with time-varying topological structure (SOM-TVS) clustering algorithm is proposed. First, we combine the locations and weights of the neurons in competition layer and initialize the SOM according to multipath components (MPCs), in which the topology of the SOM is changed with the iterations of the algorithm. Then, a new MPCs’ power weighted Gaussian–Sinc function is proposed to optimize the competitiveness performance of the algorithm. Finally, both the measured and simulated channels in millimeter wave (mmWave) are used to validate the performance of the proposed SOM-TVS algorithm. The SOM-TVS algorithm can perform great effectiveness and robustness based on numerical simulations and experimental validation results. More significantly, the proposed clustering algorithm does not require any prior knowledge of cluster and has a fairly low complexity.

Study on Identification and Prevention of Traffic Congestion Zones Considering Resilience-Vulnerability of Urban Transportation Systems

In order to solve the problem of urban short-term traffic congestion and temporal and spatial heterogeneity, it is important to scientifically delineate urban traffic congestion response areas to alleviate regional traffic congestion and improve road network efficiency. Previous urban traffic congestion zoning is mostly divided by urban administrative divisions, which is difficult to reflect the difference of congestion degree within administrative divisions or traffic congestion zoning. In this paper, we introduce the Self-Organizing Feature Mapping (SOFM) model, construct the urban traffic congestion zoning index system based on the resilience and vulnerability of urban traffic systems, and establish the urban traffic congestion zoning model, which is divided into four, five, six, and seven according to the different structures of competition layer topology. The four vulnerability damage capacity indicators of traffic volume, severe congestion mileage, delay time and average operating speed, and two resilience supply capacity indicators of traffic systems, namely, road condition and number of lanes, are used as model input vectors; the data of Guiyang city from January to June 2021 are used as data sets to input four SOFM models for training and testing and the best SOFM model with six competitive topologies is constructed. Finally, the Support Vector Machine (SVM) is used to identify the optimal partition boundary line for traffic congestion. The results show that the four models predict the urban traffic congestion zoning level correctly over 95% on the test set, each traffic congestion zoning evaluation index in the urban area shows different obvious spatial clustering characteristics, the urban traffic congestion area is divided into six categories, and the city is divided into 16 zoning areas considering the urban traffic congestion control types (prevention zone, control zone, closure control zone). The spatial boundary is clear and credible, which helps to improve the spatial accuracy when predicting urban traffic congestion zoning and provides a new methodological approach for urban traffic congestion zoning and zoning boundary delineation.

Design of Winner-Takes-All Circuits in Competitive Neural Networks

The Winner-Take-All circuit is an important part of the competition layer in the competitive neural network. Its main function is to compare the size of the output of the nodes after the weighted summation of all input vectors, and select the node with the largest output to output high power level, while other nodes output low level, that is, to find the node with the largest output. According to the characteristics of the Winner-Take-All circuit in the competitive neural network, the simulation of the Winner-Take-All circuit is carried out by the PSPICE simulation software. The physical test results show that, like the simulation diagram of the Winner-Take-All circuit, it conforms to the logic truth table, which further confirms the rationality and correctness of the Winner-Take-All circuit. Hardware realization of Winner-Take-All circuit as an important component of competitive layer in competitive neural networks has important research significance.

Multi-Associated Parameters Aggregation-Based Routing and Resources Allocation in Multi-Core Elastic Optical Networks

Space division multiplexing (SDM), as a potential means of enhancing the capacity of optical transmission systems, has attracted widespread attention. However, the adoption of SDM technology has also additionally increased resource dimensions, introduced complex crosstalk, and made it difficult to integrate multi-dimensional fragments. These factors force the transmission constraints to be more complicated. Especially, some factors have a mutual restraint relationship, and excessive consideration of certain factors will cause the deterioration of other ones. Therefore, how to comprehensively consider the associated factors to achieve trade-offs and improve network performance is a problem worthy of study. This paper exploits the advantages of self-organizing feature mapping (SOFM) model to process multi-dimensional data with relevant features. Firstly, multiple constraints will be input into SOFM as mode vectors from the core level. Then, by judging the similarity between the competition layer neuron and the pattern vector, the position of the winning neuron is located, which determines the transmission level of each core. Finally, a routing, core, and spectrum allocation scheme is proposed by preferentially locating the core with higher transmission quality. Along the selected core, the available slots will be classified twice respectively by the number of adjacent cores and crosstalk direction to quickly find the spectrum blocks with relatively small crosstalk. Results indicate the scheme can reduce blocking probability and the resource fragmentation. Further, it can increase the resource utilization within tested network load.

Transition Metal Dichalcogenides for High−Performance Aqueous Zinc Ion Batteries

Aqueous zinc ion batteries (ZIBs) with cost—effectiveness, air stability, and remarkable energy density have attracted increasing attention for potential energy storage system applications. The unique electrical properties and competitive layer spacing of transition metal dichalcogenides (TMDs) provide dramatical freedom for facilitating ion diffusion and intercalation, making TMDs suitable for ZIB cathode materials. The recently updated advance of TMDs for high−performance ZIB cathode materials have been summarized in this review. In particular, the key modification strategies of TMDs for realizing the full potential in ZIBs are highlighted. Finally, the insights for further development of TMDs as ZIB cathodes are proposed, to guide the research directions related to the design of aqueous ZIBs while approaching the theoretical performance metrics.

A neural network with competitive layers for character recognition

A structure and functioning mechanisms of a neural network with competitive layers are described. The network is intended to solve the character recognition task. The network consists of several competitive layers of neurons. Each layer is a neural network consisting of a number of neurons represented as a layer. The number of neural layers is equal to the number of recognized classes. All neural layers have one-to-one correspondence with one another and with the input raster. The neurons of every layer have mutual lateral learning connections, which weights are modified during the learning process. There is a competitive (inhibitory) relationship between all neural layers. This competitive interaction is realized by means of a “winner-take-all” (WTA) procedure which aim is to select the layer with the highest level of neural activity.Validation of the network has been done in experiments on recognition of handwritten digits of the MNIST database. The experiments have demonstrated that its error rate is few less than 2%, which is not a high result, but it is compensated by rather fast data processing and a very simple structure and functioning mechanisms.

A Novel Recognition Method of Aging Stage of Transformer Oil-Paper Insulation Using Raman Spectroscopic Recurrence Plots

Timely identification of the aging stage of the transformer can effectively ensure the safe operation of the transformer and prevent the aging fault of the transformer. This work proposes the aging diagnosis of transformer using the Raman spectroscopic recurrence plot combined with learning vector quantization (LVQ) neural network. Oil-paper insulation samples at different aging stages are obtained by accelerated thermal aging tests. Acquisition of Raman spectra of insulating oil is from 800 to 3000 cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−1</sup> and converts to the Raman spectroscopic recurrence plot. The gray-level cooccurrence matrix (GLCM) is used to extract the characteristic information of the Raman spectroscopic recurrence plot. The optimal number of neurons in competitive layer of LVQ neural network is determined by the experimental method. A transformer aging diagnosis model based on the Raman spectroscopic recurrence plot is established. The results show that the diagnostic accuracy of the proposed diagnostic model reached 95%. The Raman spectroscopic recurrence plot has more abundant aging “fingerprint” characteristic information and the proposed diagnostic method shows potential for monitoring oil-paper insulation.

A New Competitive Neural Architecture for Object Classification

In this paper, we propose a new neural architecture for object classification, made up from a set of competitive layers whose number and size are dynamically learned from training data using a two-step process that combines unsupervised and supervised learning modes. The first step consists in finding a set of one or more optimal prototypes for each of the c classes that form the training data. For this, it uses the unsupervised learning and prototype generator algorithm called fuzzy learning vector quantization (FLVQ). The second step aims to assess the quality of the learned prototypes in terms of classification results. For this, the c classes are reconstructed by assigning each object to the class represented by its nearest prototype, and the obtained results are compared to the original classes. If one or more constructed classes differ from the original ones, the corresponding prototypes are not validated and the whole process is repeated for all misclassified objects, using additional competitive layers, until no difference persists between the constructed and the original classes or a maximum number of layers is reached. Experimental results show the effectiveness of the proposed method on a variety of well-known benchmark data sets.

Implementation of Visual Clustering Strategy in Self-Organizing Map for Wear Studies Samples Printed Using FDM

In general, visual clusters are preferred over large data sets; this is an attempt to take advantage of cluster techniques to reduce the mathematical complexity of small data sets. To identify the possibility of implementing the clustering technique in a small dataset, the wear observations of PLA/Cu composite samples printed using the Fused Deposition Model (FDM) is taken into consideration. In this study, the Self Organizing Map (SOM) tool as a non-supervised Neural Network (NN) is used to visualize the data. Here, SOM combinations with vector quantification and projection are used to identify or rank the wear machinability parameters on the new composite filament printed under different FDM conditions. The competitive layer in SOM will classify the given parameters of the wear machine (vectors) at any number of dimensions may be into several groups of layer neurons. The limitation of SOM is map size which cannot exceed 1000 units of training. However, for the small data set under consideration, the extent of these limits will not affect performance. The SOM algorithm developed for the study of wear provides the outlet within the acceptable range. In addition, the linear regression analysis is carried out for the output response to measure the wear characteristics of the machining observation.

Identification of Vehicle Types Using Learning Vector Quantization Algorithm with Morphological Features

The increase in the number of vehicles every year results in traffic jams. So it is necessary to identify the type of vehicle, so that the vehicle can be arranged according to the path. This study aims to develop a system that can identify the type of vehicle using the Learning Vector Quantization (LVQ) algorithm. In order for LVQ to work well in identifying, information in the form of characteristics of the object is needed. For this reason, the LVQ algorithm is combined with morphological feature extraction using the parameters of area, circumference, eccentricity, major axis length, and minor axis length to obtain shape features. Based on the test results using a confusion matrix by calculating precision, recall and accuracy, it is obtained that the precision value is 85%, recall is 82% and accuracy is 83%. This paper shows that for vehicle identification, the combination of morphological feature extraction and LVQ algorithm produces a model that can identify vehicles based on their shape and classify classes through competitive layers that are supervised by a single layer network architecture, this makes the computational process faster and does not burden the computational process.&#x0D;

Особливості моделей нейромережевого класифікатора для розпізнавання об’єктів

In decision support subsystems for object recognition, the detection of the most probable result among those possible for a given set of features is of particular importance. For this purpose, it is appropriate to assign specific ranks to each of the resulting signals in the classification process. In this article, two models of the neural network classifier are considered, and the result of classification in the improved model is the formation of ranks for all defined classes using a new approach. So, the functionality of such a neural network classifier, in this case, was expanded due to the ranking of classes. The advanced neural network classifier has five layers — input, three hidden ones, and output layers. In the first hidden layer, the corresponding discriminant functions are formed, in the second hidden layer, the WTA competition mechanism is implemented (the winner takes all). The output layer, in which the object class ranks are formed, is built on counters in which the class ranks are gradually calculated. The third hidden layer acts as a masking layer, participating in the formation of ranks. Therefore, the introduction of two layers - masking and output in the form of counters — allows to determine the ranks of the input object in relation to its belonging to specific classes. The article presents the general structures of the considered neural network classifiers, shows the topological structures of both models of such classifiers for comparison, and also considers the functional scheme of the elements of the added layers. Features of the functioning of the proposed classifier are presented, and its structural and functional characteristics are presented in the form of a table. In addition, the peculiarities of the implementation process of the neuron competition mechanism in the competitive layer of the classifier are schematically shown.