Several graph-based methods have been proposed to perform face recognition, such as elastic graph matching, etc. These methods take advantage of the fact that the face has a graph structure. However, these methods are weaker than the CNNs. With the development of graph convolutional neural networks (GCNNs), we can reconsider the benefits of identifying the graph structure. In this paper, a face image is modeled as a sparse graph. The major challenge is how to estimate the sparse graph. Usually, the sparse graph is based on some prior clustering methods, such as k-nn, etc., that will cause the learned graph to be closer to the prior graph. Another problem is that the regularization parameters are difficult to accurately estimate. This paper presents a generic sparse graph based convolutional networks (GSgCNs). We have three advantages: 1) the regularization parameters are not estimated in the generic sparse graph modeling, 2) non-prior and 3) each sparse subgraph is represented as a connected graph of the most adjacent - relevant vertices. Because the generic sparse graph representation is non-convex, we implement the projected gradient descent algorithm with structured sparse representation. Experimental results demonstrate that the GSgCNs have good performance compared with some state-of-the-art methods.

Diabetic Retinopathy (DR) is ranked at the top of blindness causes. It progresses without subjective symptoms and leads to blindness in the worst case. However early detections and proper treatments can prevent visual disturbance. Because it takes time and cost for diagnoses by clinicians, research and development of diagnostic support systems has actively been conducted. This research aims to establish a fundus image classification method based on disease severity assessment for a diagnostic support by a fundus image analysis. In this paper, we propose a Graph Neural Network (GNN)-based method to improve accuracy for severity classification. Our method has two features. The first is to extract Region-Of-Interest (ROI) sub-images focusing on regions locally capturing lesions in order to minimize background noise in image preprocessing for the classification. The second is to utilize the GNN which is not yet applied for fundus image classification. In order to evaluate our proposed method, we use Indian Diabetic Retinopathy Image Dataset (IDRiD) utilized in "Diabetic Retinopathy: Segmentation and Grading Challenge" on Biomedical Imaging held at the IEEE International Symposium in 2018. We verified that the accuracy of our method improved 2.9% over the conventional method in this contest.

In recent years, deep learning based approaches have substantially improved the performance of face recognition. Most existing deep learning techniques work well, but neglect effective utilization of face correlation information. The resulting performance loss is noteworthy for personal appearance variations caused by factors such as illumination, pose, occlusion, and misalignment. We believe that face correlation information should be introduced to solve this network performance problem originating from by intra-personal variations. Recently, graph deep learning approaches have emerged for representing structured graph data. A graph is a powerful tool for representing complex information of the face image. In this paper, we survey the recent research related to the graph structure of Convolutional Neural Networks and try to devise a definition of graph structure included in Compressed Sensing and Deep Learning. This paper devoted to the story explain of two properties of our graph - sparse and depth. Sparse can be advantageous since features are more likely to be linearly separable and they are more robust. The depth means that this is a multi-resolution multi-channel learning process. We think that sparse graph based deep neural network can more effectively make similar objects to attract each other, the relative, different objects mutually exclusive, similar to a better sparse multi-resolution clustering. Based on this concept, we propose a sparse graph representation based on the face correlation information that is embedded via the sparse reconstruction and deep learning within an irregular domain. The resulting classification is remarkably robust. The proposed method achieves high recognition rates of 99.61% (94.67%) on the benchmark LFW (YTF) facial evaluation database.

In recent years, deep learning networks have substantially improved the performance of face recognition. Although deep learning networks have been very successful, there are limited to underlying Euclidean structure data. When dealing with complex signals such as medical imaging, genetics, social networks and computer vision, recently there has been a growing interest in trying to apply learning on non-Euclidean geometric data. Graph convolutional networks are a new deep learning architecture for analyzing non-Euclidean geometric data. In computer vision, a human face image is modeled as a graph in the irregular domain. A major technical challenge is how to optimize the structured face graph. Because, classification performance critically depends on the quality of the graph. In this paper, we explore an undirected graph convolutional network called k(3)-SGCNs (k(3)-sparse graph convolutional networks). The main idea is to use sparsity-constrained optimization that obtain connected sparse subgraphs. A sparse graph of face image is composed of connected sparse subgraphs. Experiments demonstrate that the learned sparse graph has better performance than mutual k-nearest neighbor graph and l1 graph.

Voice activity detection (VAD) is a very challenging problem in adverse acoustic environments (e.g. far-field and conditions with different types of noise). In this paper, we proposed a Gaussian mixture model (GMM) for log-energy distribution of noise and (noisy) speech, where the distribution of these two components can be self-adapting in non-stationary circumstances. An adaptive threshold based on the GMM parameters of these two components represents a reasonable bound between noise and speech, which can lead to an accurate VAD in various noise conditions. To further improve speech hit rate (SHR) and non-speech hit rate (NHR), some constraints are introduced to this proposed GMM for reliability. Experimental results demonstrate that the proposed method yields remarkable performance for SHR and NHR.

In this paper, we proposed an optimized Sparse Deep Learning Network (SDLN) model for Face Recognition (FR). A key contribution of this work is to learn feature coding of human face with a SDLN based on local structured Sparse Representation (SR). In traditional sparse FR methods, different poses and expressions of training samples could have great influence on the recognition results. We consider the SR that should be guided by context constraints which are defined by the correlations of dictionary atoms. The over-complete common dictionary that contains common atom set has been learned from a local region structured sparse encoding process. We obtained over-complete common dictionary and feature coding for each face. As we all know that the deep learning has been widely applied to face feature learning. Using traditional deep learning methods can not contain variations of face identity information. We have to get face features of compatible change in a jointly deep learning network. The proposed SDLN is jointly fine-tuned to optimize for the task of FR. The SDLN achieves high FR performance on the ORL and FERET database.

The recent years has seen immense improvement in the development of signal processing based on Curvelet transform. The Curvelet transform provide a new multi-resolution representation. The frame elements of Curvelets exhibit higher direction sensitivity and anisotropic than the Wavelets, multi-Wavelets, steerable pyramids, and so on. These features are based on the anisotropic notion of scaling. In practical instances, time series signals processing problem is often encountered. To solve this problem, the time-frequency analysis based methods are studied. However, the time-frequency analysis cannot always be trusted. Many of the new methods were proposed. The Empirical Mode Decomposition (EMD) is one of them, and widely used. The EMD aims to decompose into their building blocks functions that are the superposition of a reasonably small number of components, well separated in the time-frequency plane. And each component can be viewed as locally approximately harmonic. However, it cannot solve the problem of directionality of high-dimensional. A reallocated method of Curvelet transform (optimized Curvelet-based EMD) is proposed in this paper. We introduce a definition for a class of functions that can be viewed as a superposition of a reasonably small number of approximately harmonic components by optimized Curvelet family. We analyze this algorithm and demonstrate its results on data. The experimental results prove the effectiveness of our method.

The last decade have seen tremendous improvement in the development of new image information processing and computational tools based on sparse representation.Today, in the information sciences, computer vision and image processing, the development of sparse representation algorithms led to convenient tools to transient compressed image (data) rapidly, to remove noise from image, and to get the super-resolution image.In the study of sparse representation of images, overcomplete dictionary is used.It contains prototype imageatoms.In this way, the images are described by sparse linear combinations of theses atoms.In this field has concentrated mainly on the design of a better dictionary.The generalized K-Means algorithm (K-SVD) [1] taught us a very good case.This paper has proposed an optimization algorithm adopting the Bayesian tracking and K-SVD analysis method.We analyze this algorithm and demonstrate its results on image data.

Metal–matrix composites reinforced with sub-micrometre particles of TiC and AlN are made by in situ reaction of CH 4 and NH 3 gases with an Al–6.2Ti–4.6Mg (wt.%) melt, with a range of processing conditions being explored. High-resolution electron microscopy of the particle/matrix interfaces show that in all cases they are clean and well bonded. The orientation relationships between the various types of particle and the matrix are examined. In samples where the Ti is completely consumed by the reaction, the TiC particles are coated with a thin layer of Al 3 Ti. The presence of this layer has little effect on the mechanical properties of the composites.

Montmorillonite (MMT)/starch composites with different kinds of MMT were prepared and investigated by XRD and DMTA, and the structures, properties and their relationship were discussed. The results indicate that the properties of the composites are directly related to the dispersion of MMT′s layers. After the swelling treatment of the organic montmorillonite, the montmorillonite layers tended to be disorderly dispersed and the properties of the montmorillonite/starch composites were improved correspondingly.

综述了近年来针对支架手术中再狭窄而采用的几种金属支架表面改性的方法与可降解医用高分子支架的发展、研究现状,其中包括对目前常用的可降解材料--聚乳酸进行了概述,以及应用有限元方法对可降解支架的支撑力进行的初步探索.

The behavior of second-phase particles in a microalloyed steel during weld thermal cycle has been investigated using analytical electron microscopy. All the particles in the base metal contains Nb and Ti, and the Nb weight fraction of the analyzed particles ranged from 0.22 to 0.85. During the heating stage of the weld thermal cycle, the small particles which were rich in Nb disappeared through complete dissolution and the large particles survived the weld thermal cycle with Nb weight fraction decreased. With a high cooling rate (t8/5 no greater than 60s), only heterogeneous re-precipitation occurred during cooling. With a low cooling rate (t8/5 = 120s), homogenous re-precipitation also occurred, producing a large number of small Nb-rich particles during cooling.

Oxidation behavior of in-situ synthesized 10vol% TiC/Ti-6Al metal matrix composite was studied by using TGA at 600degreesC, 700degreesC and 800degreesC, respectively, for 20 hours. The experimental results show that oxidation of in-situ TiC/Ti-6Al composite at high temperatures basically followed a parabolic law. The activation energy for oxidation was calculated to be 255.7 kJ/mol. And the weight gain when oxidized at 800degreesC was much higher than that at 600degreesC and 700degreesC. The oxides at 800degreesC were mainly TiO2 and Al2O3, while the oxidation products contained Ti, Al, O and some carbon when oxidized at 600degreesC and 700degreesC. The microstructure of oxides revealed that the oxide formed an integrated and continuous film on the surface of sample when oxidized at 800degreesC. However, for the cases oxidized at 600degreesC and 700degreesC the oxides were un-continuous and as island-shape. The formation of oxidation islands was contributed to the higher reactivity of TiC and oxygen than that of Ti and oxygen. The non-homogeneous oxidation mechanism in the composites due to the prior oxidation of TiC was firstly observed, which is quite different from Ti alloys and other metal materials.

Metal matrix composites with the matrix of pure Mg and the hybrid reinforcements of SiC particulates and Al2O3.SiO2 (mullite) short fibers were fabricated by the liquid pressure infiltration process. It was found that the tensile strength and elastic modulus of Mg matrix composites with reinforcements of 8 vol% and 18 vol% was improved distinctly, compared with that of pure Mg. While damping capacity of the composites was decreased with increase of reinforcements due to the reduction of damping dependence on strain amplitude. And based on the TEM observation of dislocation configuration, it was concluded that the movement of dislocations in the composites was hindered by the reinforcements and thus reduced the damping capacity according to the G-L model.

A novel cost-effective in-situ technique was introduced to manufacture particulate reinforced titanium matrix composites, which was based on the self-consumable vacuum arc remelting (VAR) process. The reinforcing particulates, TiB and TiC, were in-situ synthesized in the titanium matrix via the addition of B4C and graphite powders during processing. The in-situ formed TiB and TiC particulates are fine and uniformly distributed in the matrix. Performances of the insitu composites are considerably improved, compared with that of titanium alloy counterpart both at room and high temperatures. The in-situ technique offers a promising route to produce particulate reinforced titanium matrix composites on a large scale from both technical and economic considerations.

TiC/Ti composites were produced by nonconsumable arcmelting technology utilizing the self-propagation high-temperature synthesis reaction between Ti and graphite. X-ray diffraction was used to identify the phases in the composite. The microstructures of the composites were observed by optical microscopy and transmission electron microscopy. The results show that there are two phases in the composite: TiC and Ti. Reinforcements are distributed uniformly in the matrix alloy. The interface between reinforcement and Ti alloy is very clean. There are high-density dislocations around the TiC particles due to the difference in coefficient of thermal expansion between TiC and Ti. The mechanical properties of the composites improve significantly due to the incorporation of reinforcement. The addition of Al not only strengthens the matrix alloy by solid-solution strengthening, but also improves the mechanical properties of the composites by refining the reinforcement and matrix alloy. The strengthening mechanisms mainly include the following factors: (a) bearing load by the TiC particles, (b) refinement of the grain size, and (c) intrinsic strengthening of the matrix alloy by high-density dislocations.

Woodceramics are new porous ecomaterials and have been proved to possess good damping characteristics compared with other ceramics, but their mechanical properties have been found to be relatively poor. In order to achieve good mechanical properties and good damping properties simultaneously, a woodceramics/ZK60A (WCMs/ZK60A hereafter) composite is fabricated using high-pressure vacuum infiltration technique. Microstructural analysis as well as measurement of mechanical properties and damping behavior of WCMs/ZK60A composite is made. Experimental results show that ZK60A alloy has infiltrated most of the pores and WCMs/ZK60A composite has an obvious interpenetrating network structure. Its strength, Young's modulus and damping characteristics are all much higher than those of unreinforced woodceramics. The intrinsic damping of woodceramics and dislocation damping in matrix are proposed as the main contributors of the damping of WCMs/ZK60A composite at low temperature while Interface damping is likely to be responsible for the majority at elevated temperature.

采用CVD技术制备了具有不同界面层的Cf/Al金属基复合材料,获得了一种界面层阻尼功能设计的新方法. 研究发现具有特殊界面层的Cf/Al复合材料的抗拉强度、弹性模量和阻尼性能比无界面层时都有明显增加,并且不同界面层的效果不同. 碳层对复合材料阻尼性能的提高效果最大,硅层的提高效果不如碳层,碳硅混合层的效果居中. 涂层的厚度也影响了阻尼提高的效果,较厚的碳层效果更好,这是由于提高了复合材料的阻尼应变振幅效应而产生的. 研究认为发生在界面层的微滑移是其主要的阻尼机制.

本文阐述了环氧树脂作为螺旋桨材料的特点.通过采用不同的固化体系,研究固化物与力学性能的关系以及不同固化体系的耐海水性能,结合实验结果和环氧体系自身的优点,为进一步制备螺旋桨复合材料确定了基体材料.