With the progress of deep learning research, convolutional neural networks have become the most important method in feature extraction. How to effectively classify and recognize the extracted features will directly af...With the progress of deep learning research, convolutional neural networks have become the most important method in feature extraction. How to effectively classify and recognize the extracted features will directly affect the performance of the entire network. Traditional processing methods include classification models such as fully connected network models and support vector machines. In order to solve the problem that the traditional convolutional neural network is prone to over-fitting for the classification of small samples, a CNN-TWSVM hybrid model was proposed by fusing the twin support vector machine (TWSVM) with higher computational efficiency as the CNN classifier, and it was applied to the traffic sign recognition task. In order to improve the generalization ability of the model, the wavelet kernel function is introduced to deal with the nonlinear classification task. The method uses the network initialized from the ImageNet dataset to fine-tune the specific domain and intercept the inner layer of the network to extract the high abstract features of the traffic sign image. Finally, the TWSVM based on wavelet kernel function is used to identify the traffic signs, so as to effectively solve the over-fitting problem of traffic signs classification. On GTSRB and BELGIUMTS datasets, the validity and generalization ability of the improved model is verified by comparing with different kernel functions and different SVM classifiers.展开更多
Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications.However,the various working conditions would degrade the diagnostic performance and m...Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications.However,the various working conditions would degrade the diagnostic performance and make gear fault diagnosis(GFD)more and more challenging.In this paper,a novel model parameter transfer(NMPT)is proposed to boost the performance of GFD under varying working conditions.Based on the previous transfer strategy that controls empirical risk of source domain,this method further integrates the superiorities of multi-task learning with the idea of transfer learning(TL)to acquire transferable knowledge by minimizing the discrepancies of separating hyperplanes between one specific working condition(target domain)and another(source domain),and then transferring both commonality and specialty parameters over tasks to make use of source domain samples to assist target GFD task when sufficient labeled samples from target domain are unavailable.For NMPT implementation,insufficient target domain features and abundant source domain features with supervised information are fed into NMPT model to train a robust classifier for target GFD task.Related experiments prove that NMPT is expected to be a valuable technology to boost practical GFD performance under various working conditions.The proposed methods provides a transfer learning-based framework to handle the problem of insufficient training samples in target task caused by variable operation conditions.展开更多
Breast Arterial Calcification(BAC)is a mammographic decision dissimilar to cancer and commonly observed in elderly women.Thus identifying BAC could provide an expense,and be inaccurate.Recently Deep Learning(DL)method...Breast Arterial Calcification(BAC)is a mammographic decision dissimilar to cancer and commonly observed in elderly women.Thus identifying BAC could provide an expense,and be inaccurate.Recently Deep Learning(DL)methods have been introduced for automatic BAC detection and quantification with increased accuracy.Previously,classification with deep learning had reached higher efficiency,but designing the structure of DL proved to be an extremely challenging task due to overfitting models.It also is not able to capture the patterns and irregularities presented in the images.To solve the overfitting problem,an optimal feature set has been formed by Enhanced Wolf Pack Algorithm(EWPA),and their irregularities are identified by Dense-kUNet segmentation.In this paper,Dense-kUNet for segmentation and optimal feature has been introduced for classification(severe,mild,light)that integrates DenseUNet and kU-Net.Longer bound links exist among adjacent modules,allowing relatively rough data to be sent to the following component and assisting the system in finding higher qualities.The major contribution of the work is to design the best features selected by Enhanced Wolf Pack Algorithm(EWPA),and Modified Support Vector Machine(MSVM)based learning for classification.k-Dense-UNet is introduced which combines the procedure of Dense-UNet and kU-Net for image segmentation.Longer bound associations occur among nearby sections,allowing relatively granular data to be sent to the next subsystem and benefiting the system in recognizing smaller characteristics.The proposed techniques and the performance are tested using several types of analysis techniques 826 filled digitized mammography.The proposed method achieved the highest precision,recall,F-measure,and accuracy of 84.4333%,84.5333%,84.4833%,and 86.8667%when compared to other methods on the Digital Database for Screening Mammography(DDSM).展开更多
特征迁移重在领域共有特征间学习,然而其忽略领域特有特征的判别信息,使算法的适应性受到一定的局限.针对此问题,提出了一种融合异构特征的子空间迁移学习(The subspace transfer learning algorithm integrating with heterogeneous fe...特征迁移重在领域共有特征间学习,然而其忽略领域特有特征的判别信息,使算法的适应性受到一定的局限.针对此问题,提出了一种融合异构特征的子空间迁移学习(The subspace transfer learning algorithm integrating with heterogeneous features,STL-IHF)算法.该算法将数据的特征空间看成共享和特有两个特征子空间的组合,同时基于经验风险最小框架将共享特征和特有特征共同嵌入到支持向量机(Support vector machine,SVM)的训练过程中.其在共享特征子空间上实现知识迁移的同时兼顾了领域特有的异构信息,增强了算法的适应性.模拟和真实数据集上的实验结果表明了所提方法的有效性.展开更多
现有的领域自适应方法在定义领域间分布距离时,通常仅从领域样本的整体分布上考虑,而未对带类标签的领域样本分布分别进行考虑,从而在一些具有非平衡数据集的应用领域上表现出一定的局限性.对此,在充分考虑源领域样本类信息的基础上,基...现有的领域自适应方法在定义领域间分布距离时,通常仅从领域样本的整体分布上考虑,而未对带类标签的领域样本分布分别进行考虑,从而在一些具有非平衡数据集的应用领域上表现出一定的局限性.对此,在充分考虑源领域样本类信息的基础上,基于结构风险最小化模型,提出了基于类分布的领域自适应支持向量机(Domain adaptation support vector machine based on class distribution,CDASVM),并将其拓展为可处理多源问题的多源领域自适应支持向量机(CDASVM from multiple sources,MSCDASVM),在人造和真实的非平衡数据集上的实验结果表明,所提方法只有优化或可比较的模式分类性能.展开更多
为获取居民公交出行的换乘信息,设计了一套基于多分类支持向量机(multi-class support vector machine)的公交换乘识别方法.通过融合GPS数据和公交IC卡数据获取训练样本,利用多分类支持向量机进行样本训练,选取最佳训练样本量,并采用网...为获取居民公交出行的换乘信息,设计了一套基于多分类支持向量机(multi-class support vector machine)的公交换乘识别方法.通过融合GPS数据和公交IC卡数据获取训练样本,利用多分类支持向量机进行样本训练,选取最佳训练样本量,并采用网格搜索法结合粒子优化算法对模型参数进行标定,以获取最优SVM分类模型.测试结果显示模型分类精度可达90%.以佛山市公交车GPS数据和IC卡数据对算法进行验证,并获取公交换乘量、公交换乘比例等基本换乘数据.结果表明:算法可在少样本条件下完成公交换乘识别,且分类识别精度高,尤其适用于公交线网复杂的大城市公交换乘识别,有助于在公交前期规划时进行线路布设和枢纽选址.展开更多
传统迁移学习方法通常直接利用相关领域中的数据来辅助完成当前领域的学习任务,而忽略了领域间互相学习的能力.针对此类问题,提出了一种具有协同约束的共生迁移学习方法(Collaborative Constraints based Symbiosis Transfer Learning,C...传统迁移学习方法通常直接利用相关领域中的数据来辅助完成当前领域的学习任务,而忽略了领域间互相学习的能力.针对此类问题,提出了一种具有协同约束的共生迁移学习方法(Collaborative Constraints based Symbiosis Transfer Learning,CCSTL).在协同约束的基础上,引入共生迁移机制实现领域间的交替互动学习,进而实现源领域和目标领域间的知识迁移,从而提高受训分类器的分类性能.在模拟数据和真实数据集上的实验结果表明了新算法的有效性.展开更多
文摘With the progress of deep learning research, convolutional neural networks have become the most important method in feature extraction. How to effectively classify and recognize the extracted features will directly affect the performance of the entire network. Traditional processing methods include classification models such as fully connected network models and support vector machines. In order to solve the problem that the traditional convolutional neural network is prone to over-fitting for the classification of small samples, a CNN-TWSVM hybrid model was proposed by fusing the twin support vector machine (TWSVM) with higher computational efficiency as the CNN classifier, and it was applied to the traffic sign recognition task. In order to improve the generalization ability of the model, the wavelet kernel function is introduced to deal with the nonlinear classification task. The method uses the network initialized from the ImageNet dataset to fine-tune the specific domain and intercept the inner layer of the network to extract the high abstract features of the traffic sign image. Finally, the TWSVM based on wavelet kernel function is used to identify the traffic signs, so as to effectively solve the over-fitting problem of traffic signs classification. On GTSRB and BELGIUMTS datasets, the validity and generalization ability of the improved model is verified by comparing with different kernel functions and different SVM classifiers.
基金Supported by National Natural Science Foundation of China(Grant No.51835009).
文摘Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications.However,the various working conditions would degrade the diagnostic performance and make gear fault diagnosis(GFD)more and more challenging.In this paper,a novel model parameter transfer(NMPT)is proposed to boost the performance of GFD under varying working conditions.Based on the previous transfer strategy that controls empirical risk of source domain,this method further integrates the superiorities of multi-task learning with the idea of transfer learning(TL)to acquire transferable knowledge by minimizing the discrepancies of separating hyperplanes between one specific working condition(target domain)and another(source domain),and then transferring both commonality and specialty parameters over tasks to make use of source domain samples to assist target GFD task when sufficient labeled samples from target domain are unavailable.For NMPT implementation,insufficient target domain features and abundant source domain features with supervised information are fed into NMPT model to train a robust classifier for target GFD task.Related experiments prove that NMPT is expected to be a valuable technology to boost practical GFD performance under various working conditions.The proposed methods provides a transfer learning-based framework to handle the problem of insufficient training samples in target task caused by variable operation conditions.
文摘Breast Arterial Calcification(BAC)is a mammographic decision dissimilar to cancer and commonly observed in elderly women.Thus identifying BAC could provide an expense,and be inaccurate.Recently Deep Learning(DL)methods have been introduced for automatic BAC detection and quantification with increased accuracy.Previously,classification with deep learning had reached higher efficiency,but designing the structure of DL proved to be an extremely challenging task due to overfitting models.It also is not able to capture the patterns and irregularities presented in the images.To solve the overfitting problem,an optimal feature set has been formed by Enhanced Wolf Pack Algorithm(EWPA),and their irregularities are identified by Dense-kUNet segmentation.In this paper,Dense-kUNet for segmentation and optimal feature has been introduced for classification(severe,mild,light)that integrates DenseUNet and kU-Net.Longer bound links exist among adjacent modules,allowing relatively rough data to be sent to the following component and assisting the system in finding higher qualities.The major contribution of the work is to design the best features selected by Enhanced Wolf Pack Algorithm(EWPA),and Modified Support Vector Machine(MSVM)based learning for classification.k-Dense-UNet is introduced which combines the procedure of Dense-UNet and kU-Net for image segmentation.Longer bound associations occur among nearby sections,allowing relatively granular data to be sent to the next subsystem and benefiting the system in recognizing smaller characteristics.The proposed techniques and the performance are tested using several types of analysis techniques 826 filled digitized mammography.The proposed method achieved the highest precision,recall,F-measure,and accuracy of 84.4333%,84.5333%,84.4833%,and 86.8667%when compared to other methods on the Digital Database for Screening Mammography(DDSM).
文摘现有的领域自适应方法在定义领域间分布距离时,通常仅从领域样本的整体分布上考虑,而未对带类标签的领域样本分布分别进行考虑,从而在一些具有非平衡数据集的应用领域上表现出一定的局限性.对此,在充分考虑源领域样本类信息的基础上,基于结构风险最小化模型,提出了基于类分布的领域自适应支持向量机(Domain adaptation support vector machine based on class distribution,CDASVM),并将其拓展为可处理多源问题的多源领域自适应支持向量机(CDASVM from multiple sources,MSCDASVM),在人造和真实的非平衡数据集上的实验结果表明,所提方法只有优化或可比较的模式分类性能.
文摘为获取居民公交出行的换乘信息,设计了一套基于多分类支持向量机(multi-class support vector machine)的公交换乘识别方法.通过融合GPS数据和公交IC卡数据获取训练样本,利用多分类支持向量机进行样本训练,选取最佳训练样本量,并采用网格搜索法结合粒子优化算法对模型参数进行标定,以获取最优SVM分类模型.测试结果显示模型分类精度可达90%.以佛山市公交车GPS数据和IC卡数据对算法进行验证,并获取公交换乘量、公交换乘比例等基本换乘数据.结果表明:算法可在少样本条件下完成公交换乘识别,且分类识别精度高,尤其适用于公交线网复杂的大城市公交换乘识别,有助于在公交前期规划时进行线路布设和枢纽选址.
文摘传统迁移学习方法通常直接利用相关领域中的数据来辅助完成当前领域的学习任务,而忽略了领域间互相学习的能力.针对此类问题,提出了一种具有协同约束的共生迁移学习方法(Collaborative Constraints based Symbiosis Transfer Learning,CCSTL).在协同约束的基础上,引入共生迁移机制实现领域间的交替互动学习,进而实现源领域和目标领域间的知识迁移,从而提高受训分类器的分类性能.在模拟数据和真实数据集上的实验结果表明了新算法的有效性.
