Sparse spectral hashing

A better similarity index structure for high-dimensional feature datapoints is very desirable for building scalable content-based search systems on feature-rich dataset. In this paper, we introduce sparse principal component analysis (Sparse PCA) and Boosting Similarity Sensitive Hashing (Boosting SSC) into traditional spectral hashing for both effective and data-aware binary coding for real data. We call this Sparse Spectral Hashing (SSH). SSH formulates the problem of binary coding as a thresholding a subset of eigenvectors of the Laplacian graph by constraining the number of nonzero features. The convex relaxation and eigenfunction learning are conducted in SSH to make the coding globally optimal and effective to datapoints outside the training data. The comparisons in terms of F1 score and AUC show that SSH outperforms other methods substantially over both image and text datasets.     

利用拉普拉斯能量和CNN的多聚焦图像融合方法

多聚焦图像融合技术的目的是生成一幅全聚焦图像.所谓全聚焦图像,就是将不同源图像的清晰区域集成到一幅单一的图像中.传统的图像融合方法通常存在块伪影、人造边、晕轮效果、振铃效果以及对比度下降等问题.对此,本文提出了一种利用拉普拉斯能量和CNN的多聚焦图像融合方法.使用拉普拉斯能量算子可以有效的提取源图像的聚焦信息,而训练后的卷积神经网络模型从聚焦信息图中提取的聚焦特征可以有效的进行聚焦子块和离焦子块的区分.训练后的卷积神经网络模型不仅具有很好的提取活跃窗口相对聚焦度的能力,而且可以获得精确的分割边界.在经过多轮训练后,卷积神经网络模型可以很好的在源图像和分值图之间建立一种有效的映射,这对于生成一幅精准的聚焦图至关重要.采用二值分割和小区域滤波技术来对聚焦图进行进一步的修正,获得用于融合的最终决策图.最后,根据最终决策图提供的权值,对多幅源图像进行融合形成最终的融合图像.实验结果表明,无论从视觉效果还是从定量评价方面,提出的方法均优于目前已有的其它融合方法.     

基于马尔科夫专家场的泊松噪声图像去噪方法

提出一种基于贝叶斯概率模型的泊松噪声图像去噪方法. 该方法基于贝叶斯最大后验概率模型框架,结合泊松概率分布,构建图像去噪模型. 考虑到马尔科夫随机场不能对复杂自然图像有效表征,引入高阶的马尔科夫专家场作为模型先验正则项,以表征图像自身概率分布. 利用二次惩罚函数,优化求解去噪模型,还原清晰图像. 将所提方法与其他去噪算法进行仿真实验对比,并采用峰值信噪比和结构相似性2种评价指标对去噪效果进行客观评价. 实验结果表明：与传统去噪方法相比,该方法的峰值信噪比至少提升了0.18 dB,去噪性能显著优于其他方法,能更好地保留图像的细节信息.     

Sentiment visualization and classification via semi-supervised nonlinear dimensionality reduction

Sentiment analysis, which detects the subjectivity or polarity of documents, is one of the fundamental tasks in text data analytics. Recently, the number of documents available online and offline is increasing dramatically, and preprocessed text data have more features. This development makes analysis more complex to be analyzed effectively. This paper proposes a novel semi-supervised Laplacian eigenmap (SS-LE). The SS-LE removes redundant features effectively by decreasing detection errors of sentiments. Moreover, it enables visualization of documents in perceptible low dimensional embedded space to provide a useful tool for text analytics. The proposed method is evaluated using multi-domain review data set in sentiment visualization and classification by comparing other dimensionality reduction methods. SS-LE provides a better similarity measure in the visualization result by separating positive and negative documents properly. Sentiment classification models trained over reduced data by SS-LE show higher accuracy. Overall, experimental results suggest that SS-LE has the potential to be used to visualize documents for the ease of analysis and to train a predictive model in sentiment analysis. SS-LE can also be applied to any other partially annotated text data sets.     

基于稀疏流形聚类嵌入模型和L1范数正则化的标签错误检测

综合利用含错标签中的有用信息和数据结构中蕴含的鉴别信息,提出一种基于稀疏流形聚类嵌入模型和L1范数正则化的标签错误检测修正方法.首先,用稀疏流形聚类嵌入模型将数据投影到易分类的空间,利用标注正确的极少量样本和最近邻分类器获得新标签;然后,构造标签错误检测模型,获得仅含0、1元素的检测向量,正确、错误的标签分别对应着1、0的位置;最后,给出了相应的优化算法及收敛证明,并在相关实验上验证了算法的有效性.     

基于梯度正则化联合重构河流水质模型多项参数

为识别河流水质模型参数,首先采用Crank-Nicolson有限差分格式离散控制方程,其次通过附加终端观测值,设计梯度正则化算法重构河流水质模型的多项参数。结果表明,不管是顺直河流的常系数模型,还是天然的线性相关或者线性无关的变系数模型,梯度正则化方法均能快速有效识别模型参数;初值和测量误差对参数重构影响较小,充分证明梯度正则化算法是一种有效稳定地反演河流水质模型参数的方法。     

基于拉普拉斯方向的差值线性判别分析

标准的LDA方法通常有3个问题:1)为了确保类内散度矩阵的非奇异性,必须首先通过PCA进行维数约简,这限制了对更多维数空间的使用;2)当每人只有单个训练样本时,类内散度矩阵必然奇异,此时LDA无法工作;3)缺乏对像素间的局部相关性的考虑。为了解决这些问题,提出一种基于拉普拉斯方向的差值线性判别分析方法。该方法通过拉普拉斯方向实现更鲁棒的图像相异性测度,通过引入差值散度矩阵来避免类内散度矩阵的奇异性。实验结果显示,该算法对表情变化、光照改变及不同遮挡情况获得了更高的识别率,尤其针对光照变化,效果更加显著。     

基于流形正则化的非光滑非负矩阵分解

经典的非光滑非负矩阵分解方法只能发现数据中的全局统计信息,对于非线性分布数据无能为力,而流形学习方法在探索高维非线性数据集真实几何结构方面具有明显优势。鉴于此,基于流形正则化思想,提出了一种新颖的基于流形正则化的非光滑非负矩阵分解方法。该方法不仅考虑了数据的几何结构,而且对编码系数矩阵和基矩阵同时进行稀疏约束,并将它们整合于单个目标函数中。构造了一个有效的乘积更新算法,并在理论上证明了算法的收敛性。标准数据集上的实验表明了MRnsNMF的有效性。     

脑电Laplacian空域滤波的改进算法

Laplacian空域滤波通过被关注的通道与相邻通道信号转换,能有效地提高脑电信号的信噪比,经典的Laplacian空域滤波矩阵由通道之间的距离确定.该方法依赖于相应通道定位的准确性,实际应用中不能保证电极精准安放并存在个体差异,影响了该方法的应用效果.为了提高该算法的通用性,采用高斯模型改进经典Laplacian空域滤波算法,根据用户调试数据确定模型相关参数,进而确定空域滤波矩阵.采用国际脑机接口竞赛BCI competition 2005 Ⅳ数据集验证滤波后的效果,总体比经典Laplacian方法识别准确度提高10个百分比,采用改进方法滤波后不同类间距离增大、同类间距离减小,提高了信号的信噪比.     

Laplacian group sparse modeling of human actions

Recently, many local-feature based methods have been proposed for feature learning to obtain a better high-level representation of human behavior. Most of the previous research ignores the structural information existing among local features in the same video sequences, while it is an important clue to distinguish ambiguous actions. To address this issue, we propose a Laplacian group sparse coding for human behavior representation. Unlike traditional methods such as sparse coding, our approach prefers to encode a group of relevant features simultaneously and meanwhile allow as less atoms as possible to participate in the approximation so that video-level sparsity is guaranteed. By incorporating Laplacian regularization the method is capable to ensure the similar approximation of closely related local features and the structural information is successfully preserved. Thus, a compact but discriminative human behavior representation is achieved. Besides, the objective of our model is solved with a closed-form solution, which reduces the computational cost significantly. Promising results on several popular benchmark datasets prove the efficiency and effectiveness of our approach.