Research and Improvements of ART-2 Neural Networks

唐红卫,桑农,曹治国,张天序

DOI: https://doi.org/10.3969/j.issn.1007-2276.2004.01.025

2004-01-01

Abstract:ART-2 Neural Networks are well applied in clustering, but it discard the important information of amplitude of patterns during the procession of input data. Two improved algorithms are presented after analyzing the defect of ART-2 and experimental result is reported.

What problem does this paper attempt to address?

An Enhanced ART2 Neural Network for Clustering Analysis

Jianhong Luo,Dezhao Chen

DOI: https://doi.org/10.1109/wkdd.2008.117

2008-01-01

Abstract:The adaptive resonance theory 2 (ART2) neural network exhibits several properties which can be useful in the data mining and which are lacking in most other neural networks. But ART2 has deficiencies that the categories clustered by ART2 are very mutable to slight changes in training conditions. An improved ART2 with enhanced triplex matching mechanism, named as ETM-ART2, is presented to redress the deficiencies. Several tests results show that ETM-ART2 performs better than classic ART2 when applied to clustering tasks. It is an effective improved algorithm and can be applied to a wide variety of problems.
A new ART neural networks for remote sensing image classification

AnFei Liu,BiCheng Li,Gang Chen,Xianfei Zhang

DOI: https://doi.org/10.1007/11539117_6

2005-01-01

Abstract:A new ART2A-C algorithm based on fuzzy operators to cluster the remote sensing images and aerials is proposed in this paper. By combining two ART ANNs with higher performance, the traditional ART2A-C is developed with the fuzzy operators introduced in matching rule. Then the proposed method is applied to the classification and the new network is implemented as well as two other existed ARTs respectively. Experimental results show that the new method outperforms the traditional ones.
ART2 neural network with bidirectional matching mechanism

钟旭,陈德钊,陈亚秋,罗建宏

DOI: https://doi.org/10.3785/j.issn.1008-973X.2004.12.002

2004-01-01

Abstract:A novel bidirectional matching adaptive resonance theory (BM-ART2) neural network was proposed to improve insensitivity to gradual processes of traditional ART2(T-ART2) and original-pattern ART2(OP-ART2) neural network. By means of adding two bounds, the proposed network had a bidirectional matching on the clustering of input patterns and used output of comparing layer to calculate the degree of similarity. The clustering result of a series of isoceles triangle patterns showed that BM-ART2 network was sensitive to gradual processes and can perform clustering gradual patterns better than the other two ART2 networks.
The Research on a Fractionizing and Fitting ART2 Neural Network with Supervise

徐寅林,宁新宝,黄晓林

DOI: https://doi.org/10.3321/j.issn:0372-2112.2004.10.041

2004-01-01

Tien Tzu Hsueh Pao/Acta Electronica Sinica

Abstract:ART2 neural network is applied on problems widely for pattern recognition (classification), but in many cases a pattern with a specific character is not dense together. While some patterns are interleaved and disordered in the pattern space, it is difficult to separate them by a complicated surface caused by ART2 neural network. In addition, ART2 neural network doesn't learn what the classified patterns are, it means that ART2 neural network can't indicate their property. A new ART2 neural network is provided, it uses the method of fractionizing and fitting to solve above problems. Finally, it is applied on classification of high frequency electrocardiogram characteristic parameters and enhance the validity of classification greatly.
An Algorithm of Expressway Automatic Incident Detection Based on ART2 Neural Networks

Xu-hua YANG,You-xian SUN

DOI: https://doi.org/10.3969/j.issn.0451-0712.2005.08.028

2005-01-01

Abstract:A novel algorithm of expressway automatic incident detection based on the ART2 neural networks is proposed.This algorithm uses respectively the expressway traffic flow model and ART2 neural networks as observer and classifier. The residual signals will be gotten from the comparison between the actual and the estimated values of observer. the ART2 neural networks is used to classify characteristics contained in the residuals. So, whether the traffic incident has happened can be detected . It can not only recognize known traffic incident type but also can recognize unknown traffic incident type.This algorithm is a system which can be learned at the same time of working.
A novel pattern recognition algorithm: Combining ART network with SVM to reconstruct a multi-class classifier

Anna Wang,Wenjing Yuan,Junfang Liu,Zhiguo Yu,Hua Li

DOI: https://doi.org/10.1016/j.camwa.2008.10.052

IF: 3.218

2009-01-01

Computers & Mathematics with Applications

Abstract:Based on the principle of one-against-one support vector machines (SVMs) multi-class classification algorithm, this paper proposes an extended SVMs method which couples adaptive resonance theory (ART) network to reconstruct a multi-class classifier. Different coupling strategies to reconstruct a multi-class classifier from binary SVM classifiers are compared with application to fault diagnosis of transmission line. Majority voting, a mixture matrix and self-organizing map (SOM) network are compared in reconstructing the global classification decision. In order to evaluate the method's efficiency, one-against-all, decision directed acyclic graph (DDAG) and decision-tree (DT) algorithm based SVM are compared too. The comparison is done with simulations and the best method is validated with experimental data.
An improved ART 2-A model for mixed numeric and categorical data

Xiao Han,Yahui Yang,Qingni Shen,Min Xia

DOI: https://doi.org/10.1109/ICIECS.2009.5365746

2009-01-01

Abstract:Adaptive Resonance Theory (ART) architectures are important neural networks for unsupervised clustering. ART 2-A is one version of the ART family capable of clustering both binary and numeric data. However, real-world problems usually contain categorical data that cannot be processed by ART 2-A. A simple solution is using binary encoding to preprocess categorical data. Binary encoding is a simple and straightforward approach, but it suffers from two main drawbacks: increase of dimensionality and lack of scalability. Therefore this paper proposes ART 2a-M, an improved version over ART 2-A. ART 2a-M can deal with mixed numeric and categorical data. Experiments were carried out on KDD Cup 99 data set to compare ART 2a-M with ART 2-A. Results show that not only ART 2a-M overcomes the two drawbacks of binary encoding. but also runs about 10% faster than the original one, while keeping the same accuracy. ©2009 IEEE.
Land Cover Classification in Xianghai Nature Reserve

Min Han,Lei Cheng,Xiaoliang Tang,Jiaohe Wang

2005-01-01

Abstract:This paper discussed the application of Fuzzy ARTMAP neural network in land cover classification in Xianghai Nature Reserve. The structure and algorithm of Fuzzy ARTMAP neural network were described in detail, and an automatic adjustment method of vigilance parameter was put forward to solve the problem of searching the optimum value in the selection of vigilance parameter. The results showed that the automatic adjustment method could adjust network parameter automatically, and improve the running speed of network. In comparing with maximum likelihood classification method and traditional Fuzzy ARTMAP neural network, the Fuzzy ARTMAP neural network with the functioning of automatic adjustment depended less on samples and had higher accuracy, and thus, could effectively make the classification of TM image covering Xianghai Nature Reserve. It's shown from the classification that in the Xianghai Nature Reserve, farmland covered large area while grassland and marsh were facing degradation, and hence, corresponding countermeasures should be taken to improve the eco-environment of this Reserve.
Application of an Improved Clustering Algorithm of Neural Networks in Performance Appraisal Systems

Yun Yi

DOI: https://doi.org/10.4018/jcit.304385

2022-08-05

Journal of Cases on Information Technology

Abstract:With the development of economic globalization, human resource competition has long become the key core of enterprise development and peer competition. Reasonably Formulating an enterprise’s employee performance appraisal management system and conducting standardized, fair, and just appraisal management are the basic requirements for the survival and development of an enterprise. This paper studies the application of an improved clustering algorithm based on neural network in an employee performance appraisal management system and explores its application value in the employee performance appraisal management system by using the improved ART2 clustering method that draws on leakage competition and Hebb rules. The experimental results of this paper show that the satisfaction of this system in the four aspects of integrated data management, system stability and convenience, and transparency in performance appraisal are all above 66%. This shows that this system has superior performance and good reference value.

English Else
On-line Quality Inspection in Laser Blank Welding Using ART2 Neural Network

Lei Zhang

2006-01-01

Abstract:Laser blank welding is becoming more and more important in the automotive industry and the quality of the weld is critical for a successful application.A fully automated solution is required to inspect the quality of the blanks.This paper presents a vision inspection system with a CMOS camera which uses ART2 network to inspect the defects on-line to obtain the geometry and the quality of the weld seam.The neural network ART2 has the capability of self-learning from the environment. It can distinguish the defects that have been learned before and give new outputs for new defects.So ART2 network is suitable for weld quality inspection in laser blank welding.Additionally,a CO_2 laser is used for the blank butt-welding.
A Pattern of Experiment Data Mining Based on Π Theorem Combined with ANN

GU Zheng-hua,TANG Hong-wu,DUAN Zi-bing

DOI: https://doi.org/10.3969/j.issn.1005-9865.2006.03.014

2006-01-01

Abstract:In the past artificial analysis of experimental data,those rich measured data were not utilized enough and effectively because of relatively unadvanced and simple analysis means in two aspects: 1) some physical knowledges which can be discovered from the data were not found adequately;2) some physical rules were misapprehensively judged out due to the influence of a few inaccurate data.To solve those problems,a pattern of experiment data mining is presented in this paper by combining artificial neural networks theory with π theorem.It was applied to natural analysis of greyhound's wake field,and results indicate that the pattern can not only solve the above problems but also reduce the need of physical model experiment to group test number,and gain an unlinear formula which has harmonious dimension and uniform format.The pattern will improve the level of physical model experiment in KDD(Knowledge Discovery from Database).
The Performance of Two CNN Methods in Artworks Aesthetic Feature Recognition

Jipeng Gao,Haolin Zhou,Yicheng Zhang

DOI: https://doi.org/10.1145/3383972.3383974

2020-01-01

Abstract:As the technology of computer vision become more advanced, it allows us to be able to classify the styles, genres and artists of artworks with the help of computer. However, never can we know how the convolutional neural networks(CNNs) extract and recognize those aesthetic features like objective or subjective elements. We apply two CNNs: VGG19 and ResNet-50 on different artworks. We compare the results from these networks to understand how these two networks work when they recognize the underlying features such as aesthetic feature. The dataset is obtained from.the best art-work on the world.. We selected five subsets from this dataset which have the most paintings and belong to five different artists. Meanwhile, these five artists have five different styles. We get 86.72% accuracy on the validation set by VGG19, while we get 82.31% accuracy on the validation set by ResNet-50. We then use many approaches such as kernel visualization, grad-cam heat-map, confusion matrix and style transfer to explore how those two convolutional neural networks extract the underlying features. By analyzing the results we can conclude that the CNNs actually has the ability to extract and learn the underlying features such as aesthetic features. We discover that different CNNs have different tendencies to extract the specific underlying features. Specifically, VGG19 prefers to extract subjective feature but ResNet-50 prefers to learn objective feature.
$\alpha$ DARTS Once More: Enhancing Differentiable Architecture Search by Masked Image Modeling

Bicheng Guo,Shuxuan Guo,Miaojing Shi,Peng Chen,Shibo He,Jiming Chen,Kaicheng Yu

DOI: https://doi.org/10.48550/arxiv.2211.10105

2022-01-01

Abstract: Differentiable architecture search (DARTS) has been a mainstream direction in automatic machine learning. Since the discovery that original DARTS will inevitably converge to poor architectures, recent works alleviate this by either designing rule-based architecture selection techniques or incorporating complex regularization techniques, abandoning the simplicity of the original DARTS that selects architectures based on the largest parametric value, namely $\alpha$. Moreover, we find that all the previous attempts only rely on classification labels, hence learning only single modal information and limiting the representation power of the shared network. To this end, we propose to additionally inject semantic information by formulating a patch recovery approach. Specifically, we exploit the recent trending masked image modeling and do not abandon the guidance from the downstream tasks during the search phase. Our method surpasses all previous DARTS variants and achieves state-of-the-art results on CIFAR-10, CIFAR-100, and ImageNet without complex manual-designed strategies.
Research of field theory based adaptive resonance neural network

Zhihua Zhou,Zhaoqian Chen,ShiFu Chen

2000-01-01

Ruan Jian Xue Bao/Journal of Software

Abstract:A Field Theory based adaptive resonance neural network algorithm FTART, which combines the advantages of Adaptive Resonance Theory and Field Theory, is proposed. FTART employs a unique approach to solve the conflicts between instances and extend classification regions dynamically. So that it does not need user to manually configure hidden units, and achieves fast training speed and high predictive accuracy. Moreover, a method named Statistic based Producing and Testing, which has the ability of extracting comprehensive and accurate symbolic rules from trained FTART, is proposed. Experimental results show that the symbolic rules extracted via this method can commendably describe the function of FTART.
Reinforced Interval Type-2 Fuzzy Clustering-based Neural Network Realized Through Attention-based Clustering Mechanism and Successive Learning

Shuangrong Liu,Sung-Kwun Oh,Witold Pedrycz,Bo Yang,Lin Wang,Jin Hee Yoon

DOI: https://doi.org/10.1109/tfuzz.2023.3321197

IF: 12.253

2023-01-01

IEEE Transactions on Fuzzy Systems

Abstract:A novel attention-based reinforced interval type-2 fuzzy clustering neural network (ARIT2FCN) is developed to improve the generalization performance of fuzzy clustering-based neural networks (FCNNs). Commonly, fuzzy rules in FCNNs are generated through the clustering-based rule generator. However, the generated fuzzy rules may not be able to fully describe the given data, because the clustering-based rule generator does not simultaneously consider the intra-cluster homogeneity and intercluster heterogeneity for both of data characteristics and label information when defining membership functions (MFs) of fuzzy rules. This negatively affects fuzzy rules to accurately quantify the inter-class heterogeneity and intra-class homogeneity, and degrades the performance of FCNNs. ARIT2FCN is proposed with the aid of the attention-based clustering mechanism and successive learning method. The attention-based clustering mechanism is designed to define MFs by simultaneously considering data characteristics and label information. The successive learning method is adopted to construct the desired fuzzy rules that can capture the inter-class heterogeneity and intra-class homogeneity. Moreover, L2 norm regularization is used to alleviate the overfitting effect. The performance of ARIT2FCN is evaluated on machine learning datasets with 16 comparative methods. In addition, two real-world problems are adopted to validate the effectiveness of ARIT2FCN. Experimental results demonstrate that ARIT2FCN outperforms the comparative methods, and the statistical tests also support the superiority of ARIT2FCN

computer science, artificial intelligence,engineering, electrical & electronic
ART: Artifact Removal Transformer for Reconstructing Noise-Free Multichannel Electroencephalographic Signals

Chun-Hsiang Chuang,Kong-Yi Chang,Chih-Sheng Huang,Anne-Mei Bessas

2024-09-11

Abstract:Artifact removal in electroencephalography (EEG) is a longstanding challenge that significantly impacts neuroscientific analysis and brain-computer interface (BCI) performance. Tackling this problem demands advanced algorithms, extensive noisy-clean training data, and thorough evaluation strategies. This study presents the Artifact Removal Transformer (ART), an innovative EEG denoising model employing transformer architecture to adeptly capture the transient millisecond-scale dynamics characteristic of EEG signals. Our approach offers a holistic, end-to-end denoising solution for diverse artifact types in multichannel EEG data. We enhanced the generation of noisy-clean EEG data pairs using an independent component analysis, thus fortifying the training scenarios critical for effective supervised learning. We performed comprehensive validations using a wide range of open datasets from various BCI applications, employing metrics like mean squared error and signal-to-noise ratio, as well as sophisticated techniques such as source localization and EEG component classification. Our evaluations confirm that ART surpasses other deep-learning-based artifact removal methods, setting a new benchmark in EEG signal processing. This advancement not only boosts the accuracy and reliability of artifact removal but also promises to catalyze further innovations in the field, facilitating the study of brain dynamics in naturalistic environments.

Signal Processing,Machine Learning
AN ADAPTIVE FAULT-TOLERANT NEURAL NETWORK CLASSIFIER

邵栋,周志华,陈兆乾

2001-01-01

Journal of Computer Research and Development

Abstract:Based on the research of fault-tolerant neural networks by someprecursors, an adaptive fault-tolerant neural networks learning algorithm FTART4 is proposed to resolve the multi-node open fault problem. FTART4 uses a distinct mechanism to deal with the dilemma of fault-tolerance and redundance by ascertaining a fault sensitive parameter. The problem's formalization and the specific FTART4 algorithm are given. By applying the FTART4 to some standard benchmarks, the experitments show that this method achieves preferable results.
Research on image Inpainting algorithm of improved GAN based on two-discriminations networks

Yuantao Chen,Haopeng Zhang,Linwu Liu,Xi Chen,Qian Zhang,Kai Yang,Runlong Xia,Jingbo Xie

DOI: https://doi.org/10.1007/s10489-020-01971-2

IF: 5.3

2020-11-13

Applied Intelligence

Abstract:All existing image inpainting methods based on neural network models are affected by structural distortions and blurred textures on visible connectivity, such that overfitting and overlearning phenomena can easily emerge in the image inpainting processing procedure. Accordingly, in an attempt to address the defects of image inpainting algorithm, such as long iteration time, poor adaptability and unsatisfactory repairing effects, the image inpainting algorithm of improved Generative Adversarial Networks based on deep learning method of Two-Discriminations Network has been proposed in the paper. The proposed method uses image inpainting network, global discrimination network and local discrimination network to create a fusion network to apply computational images. In the training procedure of proposed algorithm, the network of image inpainting algorithm uses similar patching method to fill the broken area in image and set it as input training objects, which greatly improves the speed and quality of image inpainting. The global discrimination network uses global structure with marginal information and feature information to judge the completed image, meaning that it comprehensively achieves visible connectivity. As local discrimination network can judge the computational images, it has also been trained with assisted feature patches found on multiple images. Furthermore, the proposed method can enhance the discriminant capability and solve the problem that the image inpainting network has easily been overfitting when the features are too concentrated and limited in number to process. Our results of designed experiments demonstrate that proposed algorithm has better adaptive capability on several image categories than those state-of-the-arts.

computer science, artificial intelligence
Automatic modulation recognition of radiation source signals based on two-dimensional data matrix and improved residual neural network

Guanghua Yi,Xinhong Hao,Xiaopeng Yan,Jian Dai,Yangtian Liu,Yanwen Han

DOI: https://doi.org/10.1016/j.dt.2023.07.004

IF: 4.035

2023-07-09

Defence Technology

Abstract:Automatic modulation recognition (AMR) of radiation source signals is a research focus in the field of cognitive radio. However, the AMR of radiation source signals at low SNRs still faces a great challenge. Therefore, the AMR method of radiation source signals based on two-dimensional data matrix and improved residual neural network is proposed in this paper. First, the time series of the radiation source signals are reconstructed into two-dimensional data matrix, which greatly simplifies the signal preprocessing process. Second, the depthwise convolution and large-size convolutional kernels based residual neural network (DLRNet) is proposed to improve the feature extraction capability of the AMR model. Finally, the model performs feature extraction and classification on the two-dimensional data matrix to obtain the recognition vector that represents the signal modulation type. Theoretical analysis and simulation results show that the AMR method based on two-dimensional data matrix and improved residual network can significantly improve the accuracy of the AMR method. The recognition accuracy of the proposed method maintains a high level greater than 90% even at −14 dB SNR.

engineering, multidisciplinary
Adaptive Resonance Theory-based Topological Clustering with a Divisive Hierarchical Structure Capable of Continual Learning

Naoki Masuyama,Narito Amako,Yuna Yamada,Yusuke Nojima,Hisao Ishibuchi

DOI: https://doi.org/10.1109/ACCESS.2022.3186479

2022-07-07

Abstract:Adaptive Resonance Theory (ART) is considered as an effective approach for realizing continual learning thanks to its ability to handle the plasticity-stability dilemma. In general, however, the clustering performance of ART-based algorithms strongly depends on the specification of a similarity threshold, i.e., a vigilance parameter, which is data-dependent and specified by hand. This paper proposes an ART-based topological clustering algorithm with a mechanism that automatically estimates a similarity threshold from the distribution of data points. In addition, for improving information extraction performance, a divisive hierarchical clustering algorithm capable of continual learning is proposed by introducing a hierarchical structure to the proposed algorithm. Experimental results demonstrate that the proposed algorithm has high clustering performance comparable with recently-proposed state-of-the-art hierarchical clustering algorithms.

Machine Learning,Neural and Evolutionary Computing

Research and Improvements of ART-2 Neural Networks

An Enhanced ART2 Neural Network for Clustering Analysis

A new ART neural networks for remote sensing image classification

ART2 neural network with bidirectional matching mechanism

The Research on a Fractionizing and Fitting ART2 Neural Network with Supervise

An Algorithm of Expressway Automatic Incident Detection Based on ART2 Neural Networks

A novel pattern recognition algorithm: Combining ART network with SVM to reconstruct a multi-class classifier

An improved ART 2-A model for mixed numeric and categorical data

Land Cover Classification in Xianghai Nature Reserve

Application of an Improved Clustering Algorithm of Neural Networks in Performance Appraisal Systems

On-line Quality Inspection in Laser Blank Welding Using ART2 Neural Network

A Pattern of Experiment Data Mining Based on Π Theorem Combined with ANN

The Performance of Two CNN Methods in Artworks Aesthetic Feature Recognition

$\alpha$ DARTS Once More: Enhancing Differentiable Architecture Search by Masked Image Modeling

Research of field theory based adaptive resonance neural network

Reinforced Interval Type-2 Fuzzy Clustering-based Neural Network Realized Through Attention-based Clustering Mechanism and Successive Learning

ART: Artifact Removal Transformer for Reconstructing Noise-Free Multichannel Electroencephalographic Signals

AN ADAPTIVE FAULT-TOLERANT NEURAL NETWORK CLASSIFIER

Research on image Inpainting algorithm of improved GAN based on two-discriminations networks

Automatic modulation recognition of radiation source signals based on two-dimensional data matrix and improved residual neural network

Adaptive Resonance Theory-based Topological Clustering with a Divisive Hierarchical Structure Capable of Continual Learning