Search

In recent years, memristive crossbar-based neuromorphic computing systems (NCS) have provided a promising solution to the acceleration of neural networks. However, stuck-at faults (SAFs) in the memristor devices significantly degrade the computing accuracy of NCS. Besides, memristors suffer from process variations, causing the deviation of actual programming resistance from its target resistance. In this paper, we propose a novel reliability-driven design framework for a memristive crossbar-based NCS in combination with general and chip-specific design optimizations. First, we design a general reliability-aware training scheme to enhance the robustness of NCS to SAFs and device variations; a dropout-inspired approach is developed to alleviate the impact of SAFs; a new weighted error function, including cross-entropy error (CEE), the l 2 -norm of weights, and the sum of squares of first-order derivatives of CEE with respect to weights, is proposed to obtain a smooth error curve, where the effects of variations are suppressed. Second, given the neural network model generated by the reliability-aware training scheme, we exploit chip-specific mapping and retraining to further reduce the computation accuracy loss incurred by SAFs. Experimental results clearly demonstrate that the proposed method can boost the computation accuracy of NCS and improve the NCS robustness.

Recent years have witnessed the great successes of deep neural network (DNN) models while deploying DNN models on hardware platforms is still challenging and widely discussed. Some works proposed dedicatedly designed accelerators for some specific DNN models, while some others proposed general-purpose deployment frameworks that can optimize the hardware configurations on various hardware platforms automatically. However, the extremely large design space and the very time-consuming on-chip tests bring great challenges to the hardware configuration optimization process. In this paper, to optimize the hardware deployment, we propose an advanced active learning framework which is composed of batch transductive experiment design (BTED) and Bootstrap-guided adaptive optimization (BAO). The BTED method generates a diverse initial configuration set filled with representative configurations. Based on the Bootstrap method and adaptive sampling, the BAO method guides the selection of hardware configurations during the searching process. To the best of our knowledge, these two methods are both introduced into general DNN deployment frameworks for the first time. We embed our advanced framework into AutoTVM, and the experimental results show that our methods reduce the model inference latency by up to 28.08% and decrease the variance of inference latency by up to 92.74%.

As an underlying technology and infrastructure, blockchain creates great value when combining with other emerging information technologies, such as Artificial Intelligence, Big Data Analysis, Cryptography Technology. Although the blockchain system has been contunuously upgraded in recent years, current blockchain still has its deficiency in terms of node control and privacy protection. First of all, the high existence rate of malicious nodes in the chain made it difficult to well purify the network environment, meanwhile the malicious nodes cannot be easily dealt with. Secondly, the correlation between nodes can still be discovered through analysis. As a result, the privacy of users’ information cannot be well protected. This paper proposes a distributed identity authentication scheme based on Zero-knowledge proof, which uses face recognition technology to ensure the uniqueness and legitimacy of identity and control the generating of malicious nodes; Zero-knowledge proof algorithm is used to encrypt and support face data to prevent user identity information from being disclosed. At the same time, the linkable ring signature algorithm is adopted to confuse the correspondence between nodes and private information to further protect users’ privacy information.

Risk visualization of power system under typhoon disaster has important scientific significance and engineering application value for power system disaster prevention and mitigation. In this paper, based on the multi-source hetero-generous information database, such as equipment operation information, meteorological information and geographic information, the damage probability models of the main power tower network based on 6 machine learning algorithms including AdaBoost iteration algorithm, GBRT (Gradient Boost Regression Tree), RF (Random Forest), LR (Logistic Regression), SVR (Support Vector Regression) and CART (Classification and Regression Tree) are established by utilizing the historical damage data of the main power tower network under typhoon disaster in a coastal city, and the error and prediction accuracy of the models are compared. Then by combining with the historical data of typhoon “Mujigae”, the predicted damage probability and risk value of each model are visualized with the geographic grid of $\mathbf{0.15}^{\circ}\times \mathbf{0.1}^{\circ}$ . The predicted effectiveness of these models is compared, and the ideal model and display index is selected.

The power systems in all parts of the world are suffering from the plaguing of the wind disaster, and many coastal cities are suffering from serious typhoons. The trip fault of the power transmission line caused by the typhoon has occurred frequently, so the risk assessment of the transmission line under typhoon disaster is increasingly important. This paper summarizes some typical cases of typhoon around the world, and concludes the main reasons for the trip of transmission lines, including windage yaw, floater, tower damage, broken line, lightning stroke and so on. In view of the reasons of the trip, the Monte Carlo method, fuzzy mathematical model, machine learning algorithm, probability model, regulation method and improved electrical geometry model are introduced. The model and so on are applied to the risk assessment method and model of transmission line trip, and the preventive measures for transmission line failure under typhoon disaster are put forward. By summarizing the relevant literature on the risk assessment of power system at home and abroad, the existing literature on the risk assessment of power system in bad weather is more, but the research on the risk assessment of transmission line trip under typhoon disaster is less. Finally, this paper looks forward the development prospect of the risk assessment of transmission line trip under typhoon disaster.

Dictionary learning which is based on the sparse coding has been frequently employed to many tasks related to remote sensing images such as classification, reconstruction and change detection. Recently, many new dictionary learning algorithms which are on an non-analytic dictionary had been proposed. Online Dictionary Learning is the famous one which can be applied to process large-scale images. But the accuracy is decreased for the strategy of updating all atoms at once. So we try to propose our approach based on the improvements of ODL algorithm. In the iterations, we reasonably select special atoms within the dictionary and then introduce the particle swarm optimization into the atom updating stage of the dictionary learning model. Experiments show that our proposed algorithm improves the performance of ODL algorithm on the accuracy of reconstruction for large-scale remote sensing images. And our method has a better effect on noise suppression.

In remote sensing applications, there are often multi-source or multi-temporal images whose different components are acquired separately. Therefore, a part of the acquired images in multi-component data can be used as priors. In this paper, the reconstruction of a remote sensing image using an auxiliary image from another sensor or another time as the reference is considered. For this application, a new compressed sensing object function with an reference image as a prior is developed. In the new model, the sparsity constraints in transform domain comes from the target image, and the gradient priors in spatial domain comes from auxiliary reference image. To optimizing the the hybrid regularization, the algorithm is based on Bregman split method. The performance of the algorithm is evaluated both qualitatively and quantitatively. The results of experiment confirm that the proposed algorithm gets higher peak signal to noise ratio (PSNR) than other approaches without reference images as priors.

Objective: This paper refers to a New Rough Set neural network based on SOFM. The model perfectly solves many problems, such as the effects of training sample size and sample quality on accuracy of artificial neural network. Besides, the new network has reduced computation and time training needed, simplified the neural network structure and improved the system speed. Method: Combining the rough set with the neural network which bases on self-organized feature map (SOFM), it has presented an architecture of rough set neural network system in this paper. The paper also designs a system flow chart and describes work principle of each part. Result: The validity of these models has been tested by practical examples. Experimental results indicate that the system not only increases the quality and rate of diagnosis, but also reduces the measure items and diagnosis costs, which makes the result visualized and it has favorable applied prospect. Conclusion: The calculation result of New Rough Set neural network based on SOFM is reliable. The new model synthesizes the advantages of rough set theory and neural network.

Purpose: To promote the development of Chinese Medicine diagnosis and treatment technology, and improve people's health, and promote the development of Chinese culture, economic and social harmony. Method: Through data analysis, put forward the related influencing factors of Modernization of Chinese Medicine diagnosis and treatment technology, complement these factors by field survey and make a evaluation of the impact of these factors. Result: Put forward the related influencing factors of Modernization of Chinese Medicine diagnosis and treatment technology and confirm the impact to Modernization of Chinese Medicine diagnosis and treatment technology of these factors and give some suggestion to promote the modernization of Chinese Medicine diagnosis and treatment technology. Conclusion: Because of having incomplete running mechanism, lacking of behavior date, devoid of experience in treatment, being naked inherent connotation, the Influencing Factors of the Chinese Medicine diagnostic and therapeutic Improvement fits for the method Grey Relational Analysis.

Objective: the Promotion of the development of Chinese Medicine (including national medicine) industry in central china, for improving people's health, promoting the Chinese culture, economic and social harmony is of great significance. Methods: make grey relational analysis for the index of Chinese Medicine (including national medicine) resource status and Chinese medical institutions operation and service status in central china in 2009. Results: the results of Grey relational analysis display the relative development of Chinese Medicine (including national medicine) industry in central china area. Then, analyze the results of Grey relation and propose suggestions for relational departments. Conclusion: The action mechanism information of the Index of Chinese Medicine (including national medicine) resource status and Chinese medical institutions operation and service status in central china are incomplete, the behavioral data is rare, and the Inherent meaning is not clear, so using the method of Grey relational analysis is very appropriate.

How to handle unbalanced datasets and how to handle high-dimensional datasets are two of the most challenging issues faced by the latest machine learning research. This article reports a study aimed at providing effective solutions to these two challenges. For handling unbalanced datasets, we proposed that a different value of the cost parameter in Support Vector Machine (SVM) is employed for each class of samples. For handling high-dimensional datasets, we resorted to Independent Components Analysis (ICA), which is a multivariate analysis algorithm, along with the conventional univariate analysis. Experimental results confirmed that the proposed approaches all together significantly improved the prediction accuracy delivered by SVM.

An extracted model of coupled bonding wires is presented. The discontinuity caused by the bonding wires would significantly affect the high-speed performance of the whole system as the speed of signal propagation is increasing. The bonding wire structures with regards to the vertical and parallel coupling schemes are analyzed in this paper while their middle interconnections are represented as the cascade network of uniform transmission-line sections. Finally, the validation is implemented by using the established models and measurements to verify the presented analysis methods.