Unsupervised Feature Selection for Latent Dirichlet Allocation

As a generative model,Latent Dirichlet Allocation Model,which lacks optimization of topics' discrimination capability focuses on how to generate data,This paper aims to improve the discrimination capability through unsupervised feature selection.Theoretical analysis shows that the discrimination capability of a topic is limited by the discrimination capability of its representative words.The discrimination capability of a word is approximated by the Information Gain of the word for topics,which is used to distinguish between "general word" and "special word" in LDA topics.Therefore,we add a constraint to the LDA objective function to let the "general words" only happen in "general topics" other than "special topics".Then a heuristic algorithm is presented to get the solution.Experiments show that this method can not only improve the information gain of topics,but also make the topics easier to understand by human.

Latent Supportive Utility of Irrelevant Attributes in Feature Selection

This paper proposed a novel feature selection method LUIFS （ latent utility of irrelevant feature selection） that not only selects the relevant features, but also targets at discovering the latent useful irrelevant attributes by measuring their supportive importance to other attributes. The method minimizes the information lost and simultaneously maximizes the final classification accuracy. The classification error rates of the LUIFS method on 16 real-life datasets from UCI machine learning repository were evaluated using the ID3, Na^ve-Bayes, and IB （instance-based classifier） learning algorithms, respectively; and compared with those of the same algorithms with no feature selection （NoFS）, feature subset selection （FSS）, and correlation-based feature selection （CFS）. The empirical results demonstrate that the LUIFS can improve the performance of learning algorithms by taking the latent relevance for irrelevant attributes into consideration, and hence including those potentially important attributes into the optimal feature subset for classification.

Multi-Scale PIIFD for Registration of Multi-Source Remote Sensing Images

This paper aims at providing multi-source remote sensing images registered in geometric space for image fusion.Focusing on the characteristics and differences of multi-source remote sensing images,a feature-based registration algorithm is implemented.The key technologies include image scale-space for implementing multi-scale properties,Harris corner detection for keypoints extraction,and partial intensity invariant feature descriptor(PIIFD)for keypoints description.Eventually,a multi-scale Harris-PIIFD image registration algorithm framework is proposed.The experimental results of fifteen sets of representative real data show that the algorithm has excellent,stable performance in multi-source remote sensing image registration,and can achieve accurate spatial alignment,which has strong practical application value and certain generalization ability.

Selective ensemble modeling based on nonlinear frequency spectral feature extraction for predicting load parameter in ball mills

Strong mechanical vibration and acoustical signals of grinding process contain useful information related to load parameters in ball mills. It is a challenge to extract latent features and construct soft sensor model with high dimensional frequency spectra of these signals. This paper aims to develop a selective ensemble modeling approach based on nonlinear latent frequency spectral feature extraction for accurate measurement of material to ball volume ratio. Latent features are first extracted from different vibrations and acoustic spectral segments by kernel partial least squares. Algorithms of bootstrap and least squares support vector machines are employed to produce candidate sub-models using these latent features as inputs. Ensemble sub-models are selected based on genetic algorithm optimization toolbox. Partial least squares regression is used to combine these sub-models to eliminate collinearity among their prediction outputs. Results indicate that the proposed modeling approach has better prediction performance than previous ones.

A process monitoring method for autoregressive-dynamic inner total latent structure projection

As a dynamic projection to latent structures(PLS)method with a good output prediction ability,dynamic inner PLS(DiPLS)is widely used in the prediction of key performance indi-cators.However,due to the oblique decomposition of the input space by DiPLS,there are false alarms in the actual industrial process during fault detection.To address the above problems,a dynamic modeling method based on autoregressive-dynamic inner total PLS(AR-DiTPLS)is proposed.The method first uses the regression relation matrix to decompose the input space orthogonally,which reduces useless information for the predic-tion output in the quality-related dynamic subspace.Then,a vector autoregressive model(VAR)is constructed for the predic-tion score to separate dynamic information and static informa-tion.Based on the VAR model,appropriate statistical indicators are further constructed for online monitoring,which reduces the occurrence of false alarms.The effectiveness of the method is verified by a Tennessee-Eastman industrial simulation process and a three-phase flow system.

Multi-source Remote Sensing Image Registration Based on Contourlet Transform and Multiple Feature Fusion

Image registration is an indispensable component in multi-source remote sensing image processing. In this paper, we put forward a remote sensing image registration method by including an improved multi-scale and multi-direction Harris algorithm and a novel compound feature. Multi-scale circle Gaussian combined invariant moments and multi-direction gray level co-occurrence matrix are extracted as features for image matching. The proposed algorithm is evaluated on numerous multi-source remote sensor images with noise and illumination changes. Extensive experimental studies prove that our proposed method is capable of receiving stable and even distribution of key points as well as obtaining robust and accurate correspondence matches. It is a promising scheme in multi-source remote sensing image registration.

Domain-specific feature elimination:multi-source domain adaptation for image classification

Multi-source domain adaptation utilizes multiple source domains to learn the knowledge and transfers it to an unlabeled target domain.To address the problem,most of the existing methods aim to minimize the domain shift by auxiliary distribution alignment objectives,which reduces the effect of domain-specific features.However,without explicitly modeling the domain-specific features,it is not easy to guarantee that the domain-invariant representation extracted from input domains contains domain-specific information as few as possible.In this work,we present a different perspective on MSDA,which employs the idea of feature elimination to reduce the influence of domain-specific features.We design two different ways to extract domain-specific features and total features and construct the domain-invariant representations by eliminating the domain-specific features from total features.The experimental results on different domain adaptation datasets demonstrate the effectiveness of our method and the generalization ability of our model.

具有潜在表示和动态图约束的多标签特征选择

针对现有嵌入式方法忽略实例相关性的潜在表示对伪标记学习的影响以及固定的图矩阵导致计算误差随迭代的加深而不断增大的问题,提出一种具有潜在表示和动态图约束的多标签特征选择方法.该方法首先利用实例相关性的潜在表示构造伪标签矩阵,并将其与线性映射和最小化伪标签与真实标签之间的Friedman范数距离相结合,从而保证伪标签与真实标签之间具有较高的相似性.其次,利用伪标签的低维流形结构构建动态图,以缓解固定图矩阵导致的随迭代深度增加计算误差的问题.在12个数据集上与7种先进方法的对比实验结果表明,该方法的整体分类性能优于现有先进方法,能较好地处理多标记特征选择问题.

BIM下的高层建筑平面凹凸不规则识别方法

在高层建筑平面审查过程中,主要通过基于ResNET(残差学习网络)的图片分类模型进行平面凹凸不规则识别,存在数据样本不均衡的问题,使得最终识别结果曲线下面积(AUC)值较低。针对此类问题,提出建筑信息模型(BIM)下的高层建筑平面凹凸不规则识别方法。利用布尔交运算方法,分析高层BIM模型中所有结合对象之间的关系,获取建筑平面信息。通过无人机搭载图像采集系统获取建筑实景图像,并应用多尺度局部直方图均衡化算法实现图像增强。建立以最大类间方差法为基础的图像分割方案,结合多策略融合未来搜索算法确定最优分割阈值,得到高层建筑图像目标前景区域。最后,运用深度学习网络构建基于潜在特征空间的图片异常检测模型,将分割后图像输入训练好的模型中自动学习,即可得到建筑平面凹凸不规则识别结果。实验结果表明:该方法识别结果的AUC值为0.9,更符合高层建筑平面审查工程的要求。

基于特征分布校准的小样本分类改进算法

针对基于特征分布校准的小样本分类算法无法准确揭示新类特征分布的问题,提出一种融合隐空间变换和密度聚类的改进算法,以解决N way-K shot任务模式下的小样本图像分类问题.首先,通过广度残差神经网络提取基类和新类图像的深度特征;其次,采用隐空间变换方法约束新类特征分布,使其更接近正态分布;再次,利用密度聚类方法为新类选取合适基类,将基类统计信息迁移到新类,并通过多元正态分布矩阵实现样本扩充;最后,构建基于集成学习的分类器,完成小样本图像分类任务.实验结果表明,相比于传统特征分布校准方法,该算法的分类准确率更高.

基于潜在特征增强网络的视频描述生成方法

视频描述生成旨在用自然语言描述视频中的物体及其相互作用。现有方法未充分利用视频中的时空语义信息,限制了模型生成准确描述语句的能力。为此,提出一种用于视频描述生成的潜在特征增强网络(LFAN)模型。利用不同的特征提取器提取外观特征、运动特征和目标特征,将对象级的目标特征分别和帧级的外观特征与运动特征融合,同时对融合后的不同特征进行增强,在生成描述前利用图神经网络和长短时记忆网络推理对象之间的时空关系,从而得到具有时空信息和语义信息的潜在特征,同时使用长短时记忆网络和门控循环单元的解码器生成视频的描述语句。该网络模型能够准确地学习到对象特征,进而引导生成更准确的词汇及与对象之间的关系。在MSVD和MSR-VTT数据集上的实验结果表明,LFAN模型可以显著提高生成描述语句的准确性,并与视频中的内容呈现出更好的语义一致性,在MSVD数据集上的BLEU@4和ROUGE-L分数分别为57.0和74.1,在MSRVTT数据集上分别为43.8和62.1。

Learning Dual-Layer User Representation for Enhanced Item Recommendation

User representation learning is crucial for capturing different user preferences,but it is also critical challenging because user intentions are latent and dispersed in complex and different patterns of user-generated data,and thus cannot be measured directly.Text-based data models can learn user representations by mining latent semantics,which is beneficial to enhancing the semantic function of user representations.However,these technologies only extract common features in historical records and cannot represent changes in user intentions.However,sequential feature can express the user’s interests and intentions that change time by time.But the sequential recommendation results based on the user representation of the item lack the interpretability of preference factors.To address these issues,we propose in this paper a novel model with Dual-Layer User Representation,named DLUR,where the user’s intention is learned based on two different layer representations.Specifically,the latent semantic layer adds an interactive layer based on Transformer to extract keywords and key sentences in the text and serve as a basis for interpretation.The sequence layer uses the Transformer model to encode the user’s preference intention to clarify changes in the user’s intention.Therefore,this dual-layer user mode is more comprehensive than a single text mode or sequence mode and can effectually improve the performance of recommendations.Our extensive experiments on five benchmark datasets demonstrate DLUR’s performance over state-of-the-art recommendation models.In addition,DLUR’s ability to explain recommendation results is also demonstrated through some specific cases.

A Two-Stage Feature Selection Method for Text Categorization by Using Category Correlation Degree and Latent Semantic Indexing

With the purpose of improving the accuracy of text categorization and reducing the dimension of the feature space,this paper proposes a two-stage feature selection method based on a novel category correlation degree(CCD)method and latent semantic indexing(LSI).In the first stage,a novel CCD method is proposed to select the most effective features for text classification,which is more effective than the traditional feature selection method.In the second stage,document representation requires a high dimensionality of the feature space and does not take into account the semantic relation between features,which leads to a poor categorization accuracy.So LSI method is proposed to solve these problems by using statistically derived conceptual indices to replace the individual terms which can discover the important correlative relationship between features and reduce the feature space dimension.Firstly,each feature in our algorithm is ranked depending on their importance of classification using CCD method.Secondly,we construct a new semantic space based on LSI method among features.The experimental results have proved that our method can reduce effectively the dimension of text vector and improve the performance of text categorization.

融合多特征和句法引导的中文命名实体识别

针对基于字符的中文命名实体识别模型中所存在一词多义和实体边界潜在词歧义的问题,提出一种融合多层语义特征和句法依存引导的中文NER模型。将句法依存引导的注意力机制与双向长短期记忆网络(BiLSTM)结合,获得字特征向量。通过迭代卷积神经网络(IDCNN)提取汉字独有特征:部首与拼音。采用协同注意力机制对句法依存引导的多种向量进行特征融合。使用CRF层来获得最佳标记序列。在多个公开数据集上的实验结果表明了模型的有效性。

基于潜在辅助特征的图像超分辨率重建算法研究

图像超分辨率重建作为图像质量增强研究领域的基本任务之一,具有很高的研究和应用价值。生成对抗网络可以有效提高超分辨率重建图像的纹理细节信息,在该领域得到了广泛应用。然而,仅仅依靠从输入的低分辨率图像中学习的特征信息,难以重建出高质量的超分辨率图像。针对该问题,本文提出一种基于潜在辅助特征的图像超分辨率重建算法,引入一个可训练的潜在特征来扩大生成器的特征空间,为重建图像提供辅助的特征信息,提高重建效果。同时还利用输入图像特征来对潜在辅助特征的生成进行约束指导,避免特征空间差异性大,导致重建图像保真度低。本文所提方法在7个公开数据集上与7种方法进行了对比实验。实验结果表明,本文方法所重建的超分图像纹理细节信息更丰富,视觉效果更好。

Estimating posterior inference quality of the relational infinite latent feature model for overlapping community detection

Overlapping community detection has become a very hot research topic in recent decades,and a plethora of methods have been proposed.But,a common challenge in many existing overlapping community detection approaches is that the number of communities K must be predefined manually.We propose a flexible nonparametric Bayesian generative model for count-value networks,which can allow K to increase as more and more data are encountered instead of to be fixed in advance.The Indian buffet process was used to model the community assignment matrix Z,and an uncol-lapsed Gibbs sampler has been derived.However,as the community assignment matrix Zis a structured multi-variable parameter,how to summarize the posterior inference results andestimate the inference quality about Z,is still a considerable challenge in the literature.In this paper,a graph convolutional neural network based graph classifier was utilized to help tosummarize the results and to estimate the inference qualityabout Z.We conduct extensive experiments on synthetic data and real data,and find that empirically,the traditional posterior summarization strategy is reliable.

融合主题特征的文本情感分析模型

随着大型语言模型的快速发展,如何在保证模型性能的同时减少模型参数量,成为了自然语言处理领的一个重要挑战。然而,现有的参数压缩技术往往难以兼顾模型的稳定性和泛化能力。为此,提出了一种融合主题特征的情感分析新架构,旨在利用主题信息增强模型对文本情感极性的判断能力。具体而言,采用一种结合LDA和K-means的方法来提取文本的主题特征,并将其作为固定维度的向量与词嵌入进行拼接,得到新的词向量表示。随后使用平均池化技术构建句子级别的表征向量,并输入到一个全连接层进行情感分类。为了验证所提模型的有效性,在公开的情感分析数据集上与多个基准算法进行了对比实验。实验结果表明,所提模型在多个数据集上明显优于ALBERT,准确率提高了约3.5%,在参数量仅有微小增加的情况下维持了较高的稳定性和泛化能力。

Posterior contraction rate of sparse latent feature models with application to proteomics

The Indian buffet process(IBP)and phylogenetic Indian buffet process(pIBP)can be used as prior models to infer latent features in a data set.The theoretical properties of these models are under-explored,however,especially in high dimensional settings.In this paper,we show that under mild sparsity condition,the posterior distribution of the latent feature matrix,generated via IBP or pIBP priors,converges to the true latent feature matrix asymptotically.We derive the posterior convergence rate,referred to as the contraction rate.We show that the convergence results remain valid even when the dimensionality of the latent feature matrix increases with the sample size,therefore making the posterior inference valid in high dimensional settings.We demonstrate the theoretical results using computer simulation,in which the parallel-tempering Markov chain Monte Carlo method is applied to overcome computational hurdles.The practical utility of the derived properties is demonstrated by inferring the latent features in a reverse phase protein arrays(RPPA)dataset under the IBP prior model.

潜伏梅毒的临床特点与血清学分析

目的总结潜伏梅毒患者的临床特点,分析其血清学检测结果。方法回顾性分析2001年1月至2007年11月在北京协和医院皮肤科性病中心诊治的601例潜伏梅毒患者的临床资料。结果 601例潜伏梅毒患者中,早期潜伏梅毒174例,晚期潜伏梅毒170例,不能确定病期的潜伏梅毒257例;男256例,女345例,男∶女为0.74∶1;20~39岁高发,主要传染来源为非婚性接触;46例(7.65%)合并其他性传播疾病;251例(41.76%)因其他性病或疑有性病进行检查时确诊。早期潜伏梅毒、晚期潜伏梅毒和不能确定病期潜伏梅毒患者中血浆反应素环状卡片试验(RPR)滴度大于1∶8的患者比例分别为72.99%(127/174)、52.94%(90/170)和60.31%(155/257)。早期潜伏梅毒患者治疗l2个月后的RPR阴转率明显低于早期显性梅毒患者(P=0.044)。结论应加强梅毒宣传力度,多渠道筛查以控制梅毒。

一种面向主题的领域服务聚类方法

随着互联网上服务资源规模的快速增长,如何高效、准确地发现服务成为一个亟待解决的关键问题.服务聚类是促进服务发现的一种重要技术.但是,现有服务聚类方法只对单一类型的服务文档进行聚类,并且没有考虑服务的领域特性.针对该问题,在对服务进行领域分类的基础上,提出了一种基于概率、融合领域特性的服务聚类模型——领域服务聚类模型(domain service clustering model,DSCM),然后基于该模型提出了一种面向主题的服务聚类方法.最后通过ProgrammableWeb网站提供的真实服务集对提出的方法进行了验证.实验结果表明,该方法可以准确地对不同类型的服务文档进行聚类.与经典的潜在狄利克雷分配(latent Dirichlet allocation,LDA),K-means等方法相比,该方法在聚类纯度和F-measure指标上均具有更好的效果,从而为按需服务发现与服务组合提供更好的支持.