摘要：As Moore＇s law based device scaling and accompanying performance scaling trends are slowing down, there is increasing interest in new technologies and computational models for fast and more energy-efficient information processing. Meanwhile, there is growing evidence that, with respect to traditional Boolean circuits and von Neumann processors, it will be challenging for beyond-CMOS devices to compete with the CMOS technology. Exploiting unique characteristics of emerging devices, especially in the context of alternative circuit and architectural paradigms, has the potential to offer orders of magnitude improvement in terms of power, performance, and capability. To take full advantage of beyond-CMOS devices, cross-layer efforts spanning from devices to circuits to architectures to algorithms are indispensable. This study examines energy-efficient neural network accelerators for embedded applications in this context. Several deep neural network accelerator designs based on cross-layer efforts spanning from alternative device technologies, circuit styles, to architectures are highlighted. Application-level benchmarking studies are presented. The discussions demonstrate that cross-layer efforts indeed can lead to orders of magnitude gain towards achieving extreme-scale energy-efficient processing.

摘要：为解决当飞行器出现破损以及飞行器受干扰时的轨迹优化问题,新一代的再入式飞行器需要有实时在线轨迹优化的能力。神经网络动态优化算法(neural dynamic optimization,NDO)的主要特点是能使神经网络逼近最优解。神经网络动态优化算法可以避免传统的间接法在求解轨迹优化问题时协态变量初值猜测问题。给出了神经网络动态优化的原理,详细介绍了优化流程。仿真结果表明神经网络动态优化算法可以很好的避免协态变量初值猜测问题,具有较强的鲁棒性,能满足实时性要求。

摘要：The paper presents the Quasi Newton model of Artificial neural Network for design of circular microstrip antenna (MSA). In this model, a closed form expression is used for accurate determination of the resonant frequency of circular microstrip patch antenna. The calculated resonant frequency results are in good agreement with the experimental results reported elsewhere. The results show better agreement with the trained and tested data of ANN models. The results are verified by the experimental results to produce accurate ANN models. This presents ANN model practically as an alternative method to the detailed electromagnetic design of circular microstrip antenna.

摘要：In this paper, we present a new design flow for robust license plate localization and recognition. The algorithm consists of three stages: 1) license plate localization;2) character segmentation;and 3) feature extraction and character recognition. The algorithm uses Mexican hat operator for edge detection and Euler number of a binary image for identifying the license plate region. A pre-processing step using median filter and contrast enhancement is employed to improve the character segmentation performance in case of low resolution and blur images. A unique feature vector comprised of region properties, projection data and reflection symmetry coefficient has been proposed. Back propagation artificial neural network classifier has been used to train and test the neural network based on the extracted feature. A thorough testing of algorithm is performed on a database with varying test cases in terms of illumination and different plate conditions. Practical considerations like existence of another text block in an image, presence of dirt or shadow on or near license plate region, license plate with rows of characters and sensitivity to license plate dimensions have been addressed. The results are encouraging with success rate of 98.10% for license plate localization and 97.05% for character recognition.

摘要：This paper, two Artificial neural Network (ANN) models using radial basis function (RBF) nets are developed for the design of Aperture Coupled Microstrip Antennas (ACMSA) for different number of design parameters. The effect of increasing the number of design parameters on the ANN model is also discussed in this work. The performances of the models when compared are found that on decreasing the number of design parameters, accuracy of the model is in-creased. The results given by the prepared models are comparable with the results of the IE3D software. So, these models are accurate enough to measure the design parameters of ACMSAs. Thus the neural network approach elimi-nates the long time consuming process of finding various designing parameters using costly software packages.

摘要：The emerging field of neuroprosthetics is focused on design and implementation of neural prostheses to restore some of the lost neural functions. Remarkable progress has been reported at most bioelectronic levels—particularly the various brain-machine interfaces (BMIs)—but the electrode-tissue contacts (ETCs) remain one of the major obstacles. The success of these BMIs relies on electrodes which are in contact with the neural tissue. Biological response to chronic implantation of Microelectrode arrays (MEAs) is an essential factor in determining a successful electrode design. By altering the material compositions and geometries of the arrays, fabrication techniques of MEAs insuring these ETCs try to obtain consistent recording signals from small groups of neurons without losing microstimulation capabilities, while maintaining low-impedance pathways for charge injection, high-charge transfer, and high-spatial resolution in recent years. So far, none of these attempts have led to a major breakthrough. Clearly, much work still needs to be done to accept a standard model of MEAs for clinical purposes. In this paper, we review different microfabrication techniques of MEAs with their advantages and drawbacks, and comment on various coating materials to enhance electrode performance. Then, we propose high-density, three-dimensional (3D), silicon-based MEAs using micromachining methods. The geometries that will be used include arrays of penetrating variable-height probes.

摘要：In this paper, a design optimization for the reinforced concrete plane frame structure has been done in order to minimize the cost of the concrete and steel for beams and columns by adopting the Artificial neural Network (ANN) computational model trough the NeuroShell-2 software program. The design procedure conforms to the ACI-318-08 Code. The variables used for design optimization are the width, depth and the area of reinforced steel, including longitudinal reinforcement and shear reinforcement. A three-bay two-story RC frame is modeled with selecting different span length and different load cases. Acceptable design results are obtained from more than 50 examples which are subjected to all the constraints of the ACI Code, using different cross-section sizes and these results are used to train the NeuroShell-2 program. The results obtained demonstrate the efficiency of the ANN procedure for the multi story RC frame design.

摘要：The key methods of detection and classification of the electroencephalogram(EEG) used in recent years are introduced . Taking EEG for example, the design plan of Kohonen neural network system based on detection and classification of complex signals is proposed, and both the network design and signal processing are analyzed, including pre-processing of signals, extraction of signal features, classification of signal and network topology, etc.

摘要：如何将优化方法和CAD建模过程无缝集成在一起是一个急需解决的有重大应用价值的课题。以蜗杆传动优化设计CAD系统的开发为例进行了初步探索。首先介绍了蜗杆传动优化设计的两大关键技术:基于BP网络实现图表的逼近,基于遗传算法求解混合离散变量优化问题。然后根据国家最新标准,建立蜗杆传动优化设计的规范化数学模型。最后介绍了Visual Studio 2008环境下采用Pro/Toolkit二次开发Pro/ENGINEER Wildfire 4.0软件开发蜗杆传动优化设计CAD系统的过程。

摘要：结合变压器分接开关内绝缘介质和场域的特点,采用ANSYS软件构建了分接开关绝缘板二维数值计算模型,分析了绝缘结构中的油隙及中心轴参数对电场分布的影响,并通过神经网络构建了绝缘板参数输入输出模型,实现了变压器动触头绝缘板结构参数最优化设计。