Predicting chaotic time series of PM10 concentration using artificial neural network

Respirable particulate matter (PM10) concentration at one residential site in Delhi, India was predicted using the neural network approach. The concepts of chaotic systems theory were utilized to build the neural network model. The embedding dimension was estimated to provide the inputs to the neural network. The model evaluation results indicated the importance of noise reduction before selecting the embedding dimension of the time series. The selection of a proper embedding dimension is considered to be essential for obtaining reliable predictions. The model’s performance shows the capability of neural networks in modelling the chaotic time series.     

Effects of Dissimilar Alumina Particulates on Microstructure and Properties of Cold-Sprayed Alumina/A380 Composite Coatings

In this study,alumina/A380 composite coatings were fabricated by cold spray.The influence of alumina particulates' morphology(spherical and irregular) and content on the deposition behavior of the coatings(including surface roughness,surface residual stress,cross-sectional microstructure and microhardness) was investigated.Results revealed that the spherical alumina mainly shows micro-tamping effect during deposition,which result in remarkable low surface roughness and porosity of the coatings.In addition,very low deposition efficiency and good interfacial bonding between the coating and the substrate were achieved.For irregular alumina particles,the embedding of ceramic particulates in the coating was dominant during deposition process,resulting in high retention in the final deposit.However,it showed limited influence on porosity,surface roughness and interfacial bonding of the deposit.The coatings containing irregular alumina particulates exhibited much higher microhardness than those containing spherical alumina due to the higher load-bearing capacity of deposited alumina.     

Construction of a pulse-coupled dipole network capable of fear-like and relief-like responses

The challenge for neuroscience as an interdisciplinary programme is the integration of ideas among the disciplines to achieve a common goal. This paper deals with the problem of deriving a pulse-coupled neural network that is capable of demonstrating behavioural responses (fear-like and relief-like). Current pulse-coupled neural networks are designed mostly for engineering applications, particularly image processing. The discovered neural network was constructed using the method of minimal anatomies approach. The behavioural response of a level-coded activity-based model was used as a reference. Although the spiking-based model and the activity-based model are of different scales, the use of model-reference principle means that the characteristics that is referenced is its functional properties. It is demonstrated that this strategy of dissection and systematic construction is effective in the functional design of pulse-coupled neural network system with nonlinear signalling. The differential equations for the elastic weights in the reference model are replicated in the pulse-coupled network geometrically. The network reflects a possible solution to the problem of punishment and avoidance. The network developed in this work is a new network topology for pulse-coupled neural networks. Therefore, the model-reference principle is a powerful tool in connecting neuroscience disciplines. The continuity of concepts and phenomena is further maintained by systematic construction using methods like the method of minimal anatomies.     

基于非线性流形学习的人脸面部运动估计

由于人脸面部运动变化分布在一个低维非线性流形中,基于线性假设的主动外观模型采用主成分分析算法描述人脸形状的变化必然带来额外的估计误差.为降低或消除这一误差,该文提出一种改进的局部线性嵌入算法构建人脸形状-纹理空间,并将其应用于主动外观模型中.实验结果表明,不仅对于面部形变不大的人脸形状,局部线性嵌入-主动外观模型拥有更...     

在PPT课件中嵌入视频文件的制作技巧

在PPT课件中使用普通方法链接的视频,如果把视频源文件删除后,链接到PPT课件中的视频就无法播放了.本文通过把视频文件转换成.swf格式大件,再利用Shockwave Flash Object控件,把.swf格式文件嵌入到PPT课件中,真正实现了视频大件的嵌入.     

基于监督局部线性嵌入的飞机目标识别方法

维数约简是目标识别的一个重要预处理步骤.由于飞机目标图像对各种空间变换(包括平移、尺度、旋转等变换)和观察角度、位置以及光照等因素都比较敏感,使得很多线性维数约简算法不能有效地用于飞机目标识别.局部线性嵌入(LLE)是一种有效的非线性维数约简方法,提出了一种基于LLE的监督LLE算法,并应用于多种条件下的飞机目标识别中...     

一种基于湿纸编码和图匹配理论的直方图保持隐写算法

该文结合隐写码、湿纸码和图匹配理论提出一种具有高嵌入效率的且可保持载体图像一阶直方图的双层嵌入隐写算法。首先构造出一条新的四进制的隐写信道,然后在该信道的LSB(Least Significant Bit)层和次LSB层分别使用二元隐写码和湿纸码嵌入秘密消息,这样即可确定出需要修改的元素,接着构造出一个图,最后使用图论中匹配算法寻找一种减小失真且保持直方图的修改方式。性能分析和实验结果表明:在嵌入相同长度秘密消息的情况下,新算法的PSNR、嵌入效率和保持一阶直方图效果均优于隐写软件Steghide。     

基于近邻元分析的半监督流形学习算法*

现有的大多数流形学习算法偏重保持流形的几何结构,并未考虑到样本点的标签信息,这在一定程度上限制了流形学习算法在数据分类中的应用.因此文中提出一种基于近邻元分析的半监督流形学习算法,采用近邻元分析学习距离度量矩阵,在距离度量方式下选择样本点的局部邻域点.基于距离度量方式构造样本点和邻域点的局部几何结构,并在样本点的低维嵌入坐标中保持这种局部几何结构不变.3个不同数据集上的分类实验验证了文中算法的有效性.