The Efect of Queueing Strategy on Network Trafc

In recent years,the transportation system has been faced by increasing challenge in congestion and inefciency,and research in trafc network has become a significant area of interest.In this paper,we introduce a dynamic-information-based(DIB)queueing strategy into network trafc model under the efcient routing strategy.DIB makes a packet with higher priority to be delivered if there are less packets travelling along its path from the current node to the destination.It is found that,compared with the traditional first-in-first-out(FIFO)queueing strategy,DIB can efectively balance the trafc load of the system via delaying packets to be delivered to congested nodes.Although the network capacity has no obvious changes,some other indexes which reflect transportation efciency are efciently improved in the congestion state.Besides,extensive simulation results and discussions are provided to explain the phenomena.The results may provide novel insights for research on trafc systems.     

A new chaotic Hopfield network with piecewise linear activation function

This paper presents a new chaotic Hopfield network with a piecewise linear activation function. The dynamic of the network is studied by virtue of the bifurcation diagram, Lyapunov exponents spectrum and power spectrum. Numerical simulations show that the network displays chaotic behaviours for some well selected parameters.     

基于双向循环神经网络的汉语语音识别*

当前基于深度神经网络模型中,虽然其隐含层可设置多层,对复杂问题适应能力强,但每层之间的节点连接是相互独立的,这种结构特性导致了在语音序列中无法利用上下文相关信息来提高识别效果,而传统的循环神经网络虽然做出了改进,但是只能对上文信息进行利用。针对以上问题,该文采用可以同时利用语音序列中上下文相关信息的双向循环神经网络模型与深度神经网络模型相结合,并应用于语音识别。构建具有5层隐含层的模型,其中第3层为双向循环神经网络结构,其他层采用深度神经网络结构。实验结果表明:加入了双向循环神经网络结构的模型与其他模型相比,较好地提高了识别正确率;噪声对双向循环神经网络汉语识别有重要影响,尤其是训练集和测试集附加噪声类型不同时,单一的含噪声语音的训练模型无法适应不同噪声类型的语音识别;调整神经网络模型中隐含层神经元数量后,识别正确率并不是一直随着隐含层中神经元数量的增加而增加,神经元数量数目增加到一定程度后正确率出现了降低的趋势。     

Observation of chimera patterns in a network of symmetric chaotic finance systems

The economic and financial systems consist of many nonlinear factors that make them behave as the complex systems. Recently many chaotic finance systems have been proposed to study the complex dynamics of finance as a noticeable problem in economics. In fact, the intricate structure between financial institutions can be obtained by using a network of financial systems. Therefore, in this paper, we consider a ring network of coupled symmetric chaotic finance systems, and investigate its behavior by varying the coupling parameters. The results show that the coupling strength and range have significant effects on the behavior of the coupled systems, and various patterns such as the chimera and multi-chimera states are observed. Furthermore, changing the parameters' values, remarkably influences on the oscillators attractors. When several synchronous clusters are formed, the attractors of the synchronized oscillators are symmetric, but different from the single oscillator attractor.     

Quantum control based on machine learning in an open quantum system

Designing robust control schemes in n-level open quantum system is significant for quantum computation. Here, we investigate two quantum control strategies based on supervised machine learning to suppress the quantum noise in an open quantum system. One is controlling state distance and the other is governing the average of a Hermitian operator. In this process, the dynamics of the system is mapped to a neural network where the control fields correspond to the weights. Besides, the system is transformed into the coherence Bloch space without using superoperator thus the complications are reduced largely. As an example, the two control protocols are demonstrated in a two-level and four-level systems, respectively. By applying these examples, the results show that the state of the system transfers to the target state and the average of a Hermitian operator to its minimum value in a given time despite disturbed by various types of noise.     

基于级联卷积神经网络的前列腺磁共振图像分类

针对深度学习训练成本高,以及基于磁共振图像的前列腺癌临床诊断需要大量医学常识且极为耗时的问题,本文提出了一种基于级联卷积神经网络（Convolutional Neural Network,CNN）和磁共振图像的前列腺癌（Prostate Cancer,PCa）自动分类诊断方法,该网络以Faster-RCNN作为前网络,对前列腺区域进行提取分割,用于排除前列腺附近组织器官的干扰;以基于ResNet改进的网络结构CNN40_bottleneck作为后网络,用于对前列腺区域病变进行分类.后网络由瓶颈结构串联组成,其中使用批量标准化（Batch Normalization,BN）、全局平均池化（Global Average Pooling,GAP）进行优化.实验结果证明,本文方法对前列腺癌诊断结果较好,而且缩减了训练时间和参数量,有效降低了训练成本.     

Aerial Video Trackers Review

Target tracking technology that is based on aerial videos is widely used in many fields; however, this technology has challenges, such as image jitter, target blur, high data dimensionality, and large changes in the target scale. In this paper, the research status of aerial video tracking and the characteristics, background complexity and tracking diversity of aerial video targets are summarized. Based on the findings, the key technologies that are related to tracking are elaborated according to the target type, number of targets and applicable scene system. The tracking algorithms are classified according to the type of target, and the target tracking algorithms that are based on deep learning are classified according to the network structure. Commonly used aerial photography datasets are described, and the accuracies of commonly used target tracking methods are evaluated in an aerial photography dataset, namely, UAV123, and a long-video dataset, namely, UAV20L. Potential problems are discussed, and possible future research directions and corresponding development trends in this field are analyzed and summarized.     

Quantum logic networks for cloning a quantum state near a given state

Two quantum logic networks are proposed to simulate a cloning machine that copies the states near a given one.Probabilistic cloning based on the first network is realized and the cloning probability of success based on the second network is 100%.Therefore,the second network is more motivative than the first one.     

Year Discrimination of Mild Aroma Chinese Liquors Using Three-Dimensional Fluorescence Spectroscopy Combined with Parallel Factor and Neural NetworkSCIEI北大核心CSCD

Three-dimensional fluorescence spectroscopy coupled with parallel factor analysis and neural network was applied to the year discrimination of mild aroma Chinese liquors. The excitation-emission fluorescence matrices (EEMs) of 120 samples with various years were measured by FLS920 fluorescence spectrometer. The trilinear decomposition of the data array was performed and the loading scores of and the excitation-emission profiles of four components were also obtained. The scores were employed as the inputs of the BP neural networks and the PARAFAC-BP identification model was constructed. 10 samples were collected from 10, 20 and 30 years of liquors respectively, and 30 samples were selected as the test sets. The remaining 90 samples were used as the training sets to build the training model. The year prediction of unknown samples was also carried out, and the prediction accuracy was 90%, 100% and 100%, respectively. Meanwhile, the discrimination analysis method and the multi way partial least squares discriminant analysis were compared, namely PARAFAC-BP and NPLS-DA. The results indicated that parallel factor combined with the neural network (PARAFAC-BP) has higher prediction accuracy. The proposed method can effectively extract the spectral characteristics, and also reduce the dimension of the input variables of neural network. A good year discrimination result was finally achieved.     

Adaptive control of chaotic systems based on a single layer neural network

This Letter presents an adaptive neural network control method for the chaos control problem. Based on a single layer neural network, the dynamic about the unstable fixed period point of the chaotic system can be adaptively identified without detailed information about the chaotic system. And the controlled chaotic system can be stabilized on the unstable fixed period orbit. Simulation results of Henon map and Lorenz system verify the effectiveness of the proposed control method.