Image classifiers that based on Deep Neural Networks(DNNs)have been proved to be easily fooled by well-designed perturbations.Previous defense methods have the limitations of requiring expensive computation or reducin...Image classifiers that based on Deep Neural Networks(DNNs)have been proved to be easily fooled by well-designed perturbations.Previous defense methods have the limitations of requiring expensive computation or reducing the accuracy of the image classifiers.In this paper,we propose a novel defense method which based on perceptual hash.Our main goal is to destroy the process of perturbations generation by comparing the similarities of images thus achieve the purpose of defense.To verify our idea,we defended against two main attack methods(a white-box attack and a black-box attack)in different DNN-based image classifiers and show that,after using our defense method,the attack-success-rate for all DNN-based image classifiers decreases significantly.More specifically,for the white-box attack,the attack-success-rate is reduced by an average of 36.3%.For the black-box attack,the average attack-success-rate of targeted attack and non-targeted attack has been reduced by 72.8%and 76.7%respectively.The proposed method is a simple and effective defense method and provides a new way to defend against adversarial samples.展开更多
Face recognition has been widely used and developed rapidly in recent years.The methods based on sparse representation have made great breakthroughs,and collaborative representation-based classification(CRC)is the typ...Face recognition has been widely used and developed rapidly in recent years.The methods based on sparse representation have made great breakthroughs,and collaborative representation-based classification(CRC)is the typical representative.However,CRC cannot distinguish similar samples well,leading to a wrong classification easily.As an improved method based on CRC,the two-phase test sample sparse representation(TPTSSR)removes the samples that make little contribution to the representation of the testing sample.Nevertheless,only one removal is not sufficient,since some useless samples may still be retained,along with some useful samples maybe being removed randomly.In this work,a novel classifier,called discriminative sparse parameter(DSP)classifier with iterative removal,is proposed for face recognition.The proposed DSP classifier utilizes sparse parameter to measure the representation ability of training samples straight-forward.Moreover,to avoid some useful samples being removed randomly with only one removal,DSP classifier removes most uncorrelated samples gradually with iterations.Extensive experiments on different typical poses,expressions and noisy face datasets are conducted to assess the performance of the proposed DSP classifier.The experimental results demonstrate that DSP classifier achieves a better recognition rate than the well-known SRC,CRC,RRC,RCR,SRMVS,RFSR and TPTSSR classifiers for face recognition in various situations.展开更多
To study the capacity of artificial neural network (ANN) applying to battlefield target classification and result of classification, according to the characteristics of battlefield target acoustic and seismic sign...To study the capacity of artificial neural network (ANN) applying to battlefield target classification and result of classification, according to the characteristics of battlefield target acoustic and seismic signals, an on the spot experiment was carried out to derive acoustic and seismic signals of a tank and jeep by special experiment system. Experiment data processed by fast Fourier transform(FFT) were used to train the ANN to distinguish the two battlefield targets. The ANN classifier was performed by the special program based on the modified back propagation (BP) algorithm. The ANN classifier has high correct identification rates for acoustic and seismic signals of battlefield targets, and is suitable for the classification of battlefield targets. The modified BP algorithm eliminates oscillations and local minimum of the standard BP algorithm, and enhances the convergence rate of the ANN.展开更多
In the process of fault detection and classification,the operation mode usually drifts over time,which brings great challenges to the algorithms.Because traditional machine learning based fault classification cannot d...In the process of fault detection and classification,the operation mode usually drifts over time,which brings great challenges to the algorithms.Because traditional machine learning based fault classification cannot dynamically update the trained model according to the probability distribution of the testing dataset,the accuracy of these traditional methods usually drops significantly in the case of covariate shift.In this paper,an importance-weighted transfer learning method is proposed for fault classification in the nonlinear multi-mode industrial process.It effectively alters the drift between the training and testing dataset.Firstly,the mutual information method is utilized to perform feature selection on the original data,and a number of characteristic parameters associated with fault classification are selected according to their mutual information.Then,the importance-weighted least-squares probabilistic classifier(IWLSPC)is utilized for binary fault detection and multi-fault classification in covariate shift.Finally,the Tennessee Eastman(TE)benchmark is carried out to confirm the effectiveness of the proposed method.The experimental result shows that the covariate shift adaptation based on importance-weight sampling is superior to the traditional machine learning fault classification algorithms.Moreover,IWLSPC can not only be used for binary fault classification,but also can be applied to the multi-classification target in the process of fault diagnosis.展开更多
针对利用海量数据构建分类模型时训练数据规模大、训练时间长且碳排放量大的问题,提出面向低能耗高性能的分类器两阶段数据选择方法TSDS(Two-Stage Data Selection)。首先,通过修正余弦相似度确定聚类中心,并将样本数据进行基于不相似...针对利用海量数据构建分类模型时训练数据规模大、训练时间长且碳排放量大的问题,提出面向低能耗高性能的分类器两阶段数据选择方法TSDS(Two-Stage Data Selection)。首先,通过修正余弦相似度确定聚类中心,并将样本数据进行基于不相似点的分裂层次聚类;其次,对聚类结果按数据分布自适应抽样以组成高质量的子样本集;最后,利用子样本集在分类模型上训练,在加速训练过程的同时提升模型精度。在Spambase、Bupa和Phoneme等6个数据集上构建支持向量机(SVM)和多层感知机(MLP)分类模型,验证TSDS的性能。实验结果表明在样本数据压缩比达到85.00%的情况下,TSDS能将分类模型准确率提升3~10个百分点,同时加速模型训练,使训练SVM分类器的能耗平均降低93.76%,训练MLP分类器的能耗平均降低75.41%。可见,TSDS在大数据场景的分类任务上既能缩短训练时间和减少能耗,又能提升分类器性能,从而助力实现“双碳”目标。展开更多
基金The work is supported by the National Key Research Development Program of China(2016QY01W0200)the National Natural Science Foundation of China NSFC(U1636101,U1736211,U1636219).
文摘Image classifiers that based on Deep Neural Networks(DNNs)have been proved to be easily fooled by well-designed perturbations.Previous defense methods have the limitations of requiring expensive computation or reducing the accuracy of the image classifiers.In this paper,we propose a novel defense method which based on perceptual hash.Our main goal is to destroy the process of perturbations generation by comparing the similarities of images thus achieve the purpose of defense.To verify our idea,we defended against two main attack methods(a white-box attack and a black-box attack)in different DNN-based image classifiers and show that,after using our defense method,the attack-success-rate for all DNN-based image classifiers decreases significantly.More specifically,for the white-box attack,the attack-success-rate is reduced by an average of 36.3%.For the black-box attack,the average attack-success-rate of targeted attack and non-targeted attack has been reduced by 72.8%and 76.7%respectively.The proposed method is a simple and effective defense method and provides a new way to defend against adversarial samples.
基金Project(2019JJ40047)supported by the Hunan Provincial Natural Science Foundation of ChinaProject(kq2014057)supported by the Changsha Municipal Natural Science Foundation,China。
文摘Face recognition has been widely used and developed rapidly in recent years.The methods based on sparse representation have made great breakthroughs,and collaborative representation-based classification(CRC)is the typical representative.However,CRC cannot distinguish similar samples well,leading to a wrong classification easily.As an improved method based on CRC,the two-phase test sample sparse representation(TPTSSR)removes the samples that make little contribution to the representation of the testing sample.Nevertheless,only one removal is not sufficient,since some useless samples may still be retained,along with some useful samples maybe being removed randomly.In this work,a novel classifier,called discriminative sparse parameter(DSP)classifier with iterative removal,is proposed for face recognition.The proposed DSP classifier utilizes sparse parameter to measure the representation ability of training samples straight-forward.Moreover,to avoid some useful samples being removed randomly with only one removal,DSP classifier removes most uncorrelated samples gradually with iterations.Extensive experiments on different typical poses,expressions and noisy face datasets are conducted to assess the performance of the proposed DSP classifier.The experimental results demonstrate that DSP classifier achieves a better recognition rate than the well-known SRC,CRC,RRC,RCR,SRMVS,RFSR and TPTSSR classifiers for face recognition in various situations.
文摘To study the capacity of artificial neural network (ANN) applying to battlefield target classification and result of classification, according to the characteristics of battlefield target acoustic and seismic signals, an on the spot experiment was carried out to derive acoustic and seismic signals of a tank and jeep by special experiment system. Experiment data processed by fast Fourier transform(FFT) were used to train the ANN to distinguish the two battlefield targets. The ANN classifier was performed by the special program based on the modified back propagation (BP) algorithm. The ANN classifier has high correct identification rates for acoustic and seismic signals of battlefield targets, and is suitable for the classification of battlefield targets. The modified BP algorithm eliminates oscillations and local minimum of the standard BP algorithm, and enhances the convergence rate of the ANN.
文摘In the process of fault detection and classification,the operation mode usually drifts over time,which brings great challenges to the algorithms.Because traditional machine learning based fault classification cannot dynamically update the trained model according to the probability distribution of the testing dataset,the accuracy of these traditional methods usually drops significantly in the case of covariate shift.In this paper,an importance-weighted transfer learning method is proposed for fault classification in the nonlinear multi-mode industrial process.It effectively alters the drift between the training and testing dataset.Firstly,the mutual information method is utilized to perform feature selection on the original data,and a number of characteristic parameters associated with fault classification are selected according to their mutual information.Then,the importance-weighted least-squares probabilistic classifier(IWLSPC)is utilized for binary fault detection and multi-fault classification in covariate shift.Finally,the Tennessee Eastman(TE)benchmark is carried out to confirm the effectiveness of the proposed method.The experimental result shows that the covariate shift adaptation based on importance-weight sampling is superior to the traditional machine learning fault classification algorithms.Moreover,IWLSPC can not only be used for binary fault classification,but also can be applied to the multi-classification target in the process of fault diagnosis.
文摘针对利用海量数据构建分类模型时训练数据规模大、训练时间长且碳排放量大的问题,提出面向低能耗高性能的分类器两阶段数据选择方法TSDS(Two-Stage Data Selection)。首先,通过修正余弦相似度确定聚类中心,并将样本数据进行基于不相似点的分裂层次聚类;其次,对聚类结果按数据分布自适应抽样以组成高质量的子样本集;最后,利用子样本集在分类模型上训练,在加速训练过程的同时提升模型精度。在Spambase、Bupa和Phoneme等6个数据集上构建支持向量机(SVM)和多层感知机(MLP)分类模型,验证TSDS的性能。实验结果表明在样本数据压缩比达到85.00%的情况下,TSDS能将分类模型准确率提升3~10个百分点,同时加速模型训练,使训练SVM分类器的能耗平均降低93.76%,训练MLP分类器的能耗平均降低75.41%。可见,TSDS在大数据场景的分类任务上既能缩短训练时间和减少能耗,又能提升分类器性能,从而助力实现“双碳”目标。
