针对当前监督学习算法在流形数据集上分类性能的缺陷,如分类精度低且稀疏性有限,本文在稀疏贝叶斯方法和流行正则化框架的基础上,提出一种稀疏流形学习算法(Manifold Learning Based on Sparse Bayesian Approach,MLSBA).该算法是对稀...针对当前监督学习算法在流形数据集上分类性能的缺陷,如分类精度低且稀疏性有限,本文在稀疏贝叶斯方法和流行正则化框架的基础上,提出一种稀疏流形学习算法(Manifold Learning Based on Sparse Bayesian Approach,MLSBA).该算法是对稀疏贝叶斯模型的扩展,通过在模型的权值上定义稀疏流形先验,有效利用了样本数据的流形信息,提高了算法的分类准确率.在多种数据集上进行实验,结果表明:MLSBA不仅在流形数据集上取得良好的分类性能,而且在非流形数据集上效果也比较好;同时算法在两类数据集上均具有良好的稀疏性能.展开更多
The all traditional electrical resistance tomography (ERT) sensors have a static structure, which cannot satisfy the intelligent requirements for adaptive optimization to ERT sensors that is subject to flow pattern ch...The all traditional electrical resistance tomography (ERT) sensors have a static structure, which cannot satisfy the intelligent requirements for adaptive optimization to ERT sensors that is subject to flow pattern changes during the real-time detection of two-phase flow. In view of this problem, an adaptive ERT sensor with a dynamic structure is proposed. The electrodes of the ERT sensor are arranged in an array structure, the flow pattern recognition technique is introduced into the ERT sensor design and accordingly an ERT flow pattern recognition method based on signal sparsity is proposed. This method uses the sparse representation of the signal to express the sampling voltage of the ERT system as a sparse combination and find its sparse solution to achieve the classification of different flow patterns. With the introduction of flow identification information, the sensor has an intelligent function of adaptively and dynamically adapting the sensor structure according to the real-time flow pattern change. The experimental results show that the sensor can automatically identify four typical flow patterns: core flow, bubble flow, laminar flow and circulation flow with recognition rates of 91%, 93%, 90% and 88% respectively. For different flow patterns, the dynamically optimized sensor can significantly improve the quality of ERT image reconstruction.展开更多
Sparse representation is a mathematical model for data representation that has proved to be a powerful tool for solving problems in various fields such as pattern recognition, machine learning, and computer vision. As...Sparse representation is a mathematical model for data representation that has proved to be a powerful tool for solving problems in various fields such as pattern recognition, machine learning, and computer vision. As one of the building blocks of the sparse representation method, dictionary learning plays an important role in the minimization of the reconstruction error between the original signal and its sparse representation in the space of the learned dictionary. Although using training samples directly as dictionary bases can achieve good performance, the main drawback of this method is that it may result in a very large and inef- ficient dictionary due to noisy training instances. To obtain a smaller and more representative dictionary, in this paper, we propose an approach called Laplacian sparse dictionary （LSD） learning. Our method is based on manifold learning and double sparsity. We incorporate the Laplacian weighted graph in the sparse representation model and impose the 11-norm sparsity on the dictionary. An LSD is a sparse overcomplete dictionary that can preserve the intrinsic structure of the data and learn a smaller dictionary for each class. The learned LSD can be easily integrated into a classification framework based on sparse representation. We compare the proposed method with other methods using three benchmark-controlled face image databases, Extended Yale B, ORL, and AR, and one uncontrolled person image dataset, i-LIDS-MA. Results show the advantages of the proposed LSD algorithm over state-of-the-art sparse representation based classification methods.展开更多
文摘针对当前监督学习算法在流形数据集上分类性能的缺陷,如分类精度低且稀疏性有限,本文在稀疏贝叶斯方法和流行正则化框架的基础上,提出一种稀疏流形学习算法(Manifold Learning Based on Sparse Bayesian Approach,MLSBA).该算法是对稀疏贝叶斯模型的扩展,通过在模型的权值上定义稀疏流形先验,有效利用了样本数据的流形信息,提高了算法的分类准确率.在多种数据集上进行实验,结果表明:MLSBA不仅在流形数据集上取得良好的分类性能,而且在非流形数据集上效果也比较好;同时算法在两类数据集上均具有良好的稀疏性能.
基金Projects(51405381,51674188)supported by the National Natural Science Foundation of China
文摘The all traditional electrical resistance tomography (ERT) sensors have a static structure, which cannot satisfy the intelligent requirements for adaptive optimization to ERT sensors that is subject to flow pattern changes during the real-time detection of two-phase flow. In view of this problem, an adaptive ERT sensor with a dynamic structure is proposed. The electrodes of the ERT sensor are arranged in an array structure, the flow pattern recognition technique is introduced into the ERT sensor design and accordingly an ERT flow pattern recognition method based on signal sparsity is proposed. This method uses the sparse representation of the signal to express the sampling voltage of the ERT system as a sparse combination and find its sparse solution to achieve the classification of different flow patterns. With the introduction of flow identification information, the sensor has an intelligent function of adaptively and dynamically adapting the sensor structure according to the real-time flow pattern change. The experimental results show that the sensor can automatically identify four typical flow patterns: core flow, bubble flow, laminar flow and circulation flow with recognition rates of 91%, 93%, 90% and 88% respectively. For different flow patterns, the dynamically optimized sensor can significantly improve the quality of ERT image reconstruction.
基金Project supported by the National Natural Science Foundation of China （Nos. 61272304 and 61363029） and the Guangxi Key Laboratory of Trusted Software （No. kx201313）
文摘Sparse representation is a mathematical model for data representation that has proved to be a powerful tool for solving problems in various fields such as pattern recognition, machine learning, and computer vision. As one of the building blocks of the sparse representation method, dictionary learning plays an important role in the minimization of the reconstruction error between the original signal and its sparse representation in the space of the learned dictionary. Although using training samples directly as dictionary bases can achieve good performance, the main drawback of this method is that it may result in a very large and inef- ficient dictionary due to noisy training instances. To obtain a smaller and more representative dictionary, in this paper, we propose an approach called Laplacian sparse dictionary （LSD） learning. Our method is based on manifold learning and double sparsity. We incorporate the Laplacian weighted graph in the sparse representation model and impose the 11-norm sparsity on the dictionary. An LSD is a sparse overcomplete dictionary that can preserve the intrinsic structure of the data and learn a smaller dictionary for each class. The learned LSD can be easily integrated into a classification framework based on sparse representation. We compare the proposed method with other methods using three benchmark-controlled face image databases, Extended Yale B, ORL, and AR, and one uncontrolled person image dataset, i-LIDS-MA. Results show the advantages of the proposed LSD algorithm over state-of-the-art sparse representation based classification methods.
