一种基于TCP的实时视频传输延时预测模型*

针对TCP重传易导致延时抖动大、降低流媒体播放质量的问题,提出了使用TCP传输实时视频时,播放质量不受TCP重传影响的必要条件,即视频帧发送延时、播放缓冲区延时与网络往返时间需要满足一定的关系式。建立了视频帧发送延时的预测模型,该模型通过输入视频帧长度、丢包率、网络往返时间、TCP拥塞窗口大小与TCP超时重传时间,预测视频帧的发送延时。NS2模拟结果表明,可以通过发送延时的预测值来判断视频帧是否适合采用TCP传输。     

Neural network modelling for environmental prediction

We describe the choice and assessment of neural network and statistical methods for data modelling, feature selection and forecasting. We deal in particular with how empirical environmental and Earth observation data can be used in conjunction with physical simulation models.     

光纤环网中建立数据传输模型的方法研究*

为了在光线环网中高效地传输数据,首先确定了所研究光纤环网的物理拓扑结构,是一种多信道单向环网,然后详细地阐述了在其中建立数据传输模型的步骤和方法,以及各个模块的流程图,最后在设计的实验平台上对模型进行了完整的测试,测试结果表明数据传输模型可以在光纤环网中高效且正确地运行。     

基于灰狼群智能最优化的神经网络PM2.5浓度预测

针对目前PM_2.5浓度测量成本高和测量过程繁杂等问题,建立了基于灰狼群智能最优化算法的神经网络预测模型。从非机理模型的角度,结合气象因素和空气污染物对上海市的PM_2.5浓度进行预测,并使用平均影响值分析了影响PM_2.5浓度的重要因素。使用灰狼群智能算法优化神经网络的过程中,引入了检验样本实时监控训练过程以避免发生"过训练"现象,确保建立的神经网络模型具有较好的泛化能力。实验结果表明：PM₁₀对PM_2.5的影响最为显著,其次是CO和前一天PM_2.5。选取2016年11月1日-12日的数据进行验证,其平均相对误差为13.46%,平均绝对误差为8μg/m³,与粒子群算法优化的神经网络、BP神经网络模型及支持向量回归（SVR）模型的误差相比,平均相对误差分别下降了约3个百分点、5个百分点和1个百分点。因此,使用灰狼算法优化的神经网络更适合上海市PM_2.5浓度的预测和空气质量的预报。     

Analytical model of data transmission in the IEEE 802.16 network

In the wireless networks controlled by the IEEE 802.16 protocol, the subscriber stations reserve the common channel using the mechanism of competitive access. Developed was an analytical model for studying the efficiency of transmission of the upward traffic in IEEE 802.16 network including the processes of channel reservation by the algorithm of multiple random access and packet transmission.     

The model of servicing the real-time traffic and data with a dynamically changeable transmission speed

A model is built up of the joint transmission of the traffic of real-time services and the data traffic admitting a delay. The real-time traffic has the advantage in the occupation and use of the channel resource. It is expressed in terms of a decrease of the data transmission speed to a certain preassigned minimum value. As a free channel resource appears, the data transfer speed increases. The scheme is investigated of the occupation of the channel resource and formal definitions are given of the basic quality indices of the joint service of requests. A solution is considered of the problem for estimating the data traffic volume that can be transmitted together with the traffic of real-time services in the fulfillment of the prescribed restrictions on the delay.     

Neural network prediction model for the methane fraction in biogas from field-scale landfill bioreactors

In this study we present a neural network model for predicting the methane fraction in landfill gas originating from field-scale landfill bioreactors. Landfill bioreactors were constructed at the Odayeri Sanitary Landfill, Istanbul, Turkey, and operated with (C2) and without (C1) leachate recirculation. The refuse height of the test cell was 5 m, with a placement area of 1250 m² (25 m × 50 m). We monitored the leachate and landfill gas components for 34 months, after which we modeled the methane fraction in landfill gas from the bioreactors (C1 and C2) using artificial neural networks; leachate components were used as input parameters. To predict the methane fraction in landfill gas as a final product of anaerobic digestion, we used input parameters such as pH, alkalinity, Chemical Oxygen Demand, sulfate, conductivity, chloride and waste temperature. We evaluated the anaerobic conversion efficiencies based on leachate characteristics during different time periods. We determined the optimal architecture of the neural network, and advantages, disadvantages and further developments of the network are discussed.     

基于变分推理回声状态网络集成模型的工业数据区间预测

针对含噪声工业时间序列数据的区间预测问题,本文提出了一种使用变分推理来数值求解回声状态网络(echo state network,ESN)集成模型参数的区间预测方法.在模型构建上,将ESN集成模型中各个ESN单元输出权值向量的先验分布方差设置为相互独立形式,相比较非独立形式更有利于模型稳定性;在模型参数求解上,本文提出用变分推理来近似推导出集成模型中所有不确定参数的联合后验概率分布,以分布中的参数均值作为模型参数值,相对于已有ESN集成模型使用最大化边缘似然度的参数估计方法效果更好.为验证提出方法的有效性,测试了人工数据集和钢铁企业真实煤气数据集.实验结果表明本文方法参数估计更为准确,在预测精度,区间质量和模型稳定性以及耗时方面表现优秀.     

面向星上实时目标检测的神经网络压缩方法

针对深度神经网络在轨实时目标检测需求与星上有限硬件资源之间的矛盾,基于宇航级处理芯片特性,提出一种结合剪枝、渐进式混合量化的混合压缩方法.对网络卷积层进行滤波器级剪枝,大幅度降低卷积计算操作数量;提出一种渐进式混合量化方法,在网络量化过程中混合使用不同的量化位宽,有效降低目标检测网络在宇航级处理芯片中的内存大小.在RSOD遥感图像数据集上的实验结果表明,该方法在网络检测精度损失<1％ 的前提下,提升了网络对星上有限计算和存储资源的利用率.     

基于差异演化算法的BP神经网络预测模型及其应用

人工神经网络的结构设计没有系统的规律可遵循,而常用的基于梯度的神经网络参数优化又易陷入局部最优解。针对BP人工神经网络所存在的缺陷,结合差异演化算法,提出了实数编码的DE-BP神经网络预测模型。利用税收预测的实例验证了算法的有效性,取得了令人满意的结果。     

Reliability analysis of a data transmission model for multicast

As computer systems have been widely used, Internet, which is a network-of-network, has been greatly developed and rapidly spread all over the world. In addition to unicast transmissions of point-to-point, multicast transmissions of one-to-many and many-to-many have been recently used. This paper considers a stochastic model where data are transmitted through multicast. The data are divided into n packets at a transmission server, and their packets are transmitted successively to each receiver of multicast group. Positive (ACK) or negative (NAK) acknowledgements are returned from each receiver, by depending on whether he or she has received all packets, correctly or not. When the server has received NAK, retransmissions only for receivers with NAK are made. The mean time until the data transmission succeeds is analytically derived and an optimal policy which maximizes the throughput is discussed. Finally, numerical examples are given under suitable conditions.     

Neural network controller for a multivariable model of submarine dynamics

Recently, there have been many attempts to use neural networks as a feedback controller. However, most of the reported cases seek to control Single-Input Single-Output (SISO) systems using some sort of adaptive strategy. In this paper, we demonstrate that neural networks can be used for the control of complex multivariable, rather than simply SISO, systems. A modified direct control scheme using a neural network architecture is used with backpropagation as the adaptive algorithm. The proposed algorithm is designed for Multi-Input Multi-Output (MIMO) systems, and is similar to that proposed by Saerens and Soquet [1] and Goldenthal and Farrell [2] for (SISO) systems, and differs only in the form of the gradient approximation. As an example of the application of this approach, we investigate the control of the dynamics of a submarine vehicle with four inputs and four outputs, in which the differential stern, bow and rudder control surfaces are dynamically coordinated to cause the submarine to follow commanded changes in roll, yaw rate, depth rate and pitch attitude. Results obtained using this scheme are compared with those obtained using optimal linear quadratic control.     

Intra-body data transmission for the personal area network

Intra-body communication is a wireless means of exchanging information within a personal area network (PAN) between wearable electronic sensors and devices. The feasibility of intra-body communication is confirmed through several experiments on signal propagation within the human body, and a human phantom is designed and used to obtain reproducible results over repeated experiments. Based on the results of these experiments, a prototype transmission system is constructed using aluminum electrodes powered by 3 V DC and operating in the 10.7 MHz frequency modulation (FM) band. This prototype is demonstrated to be capable of transmitting analog signals through the human subjects in the presence of external noise. Digital data transmission at 9600 bps is also achieved using newly fabricated 10.7 MHz frequency shift keying (FSK) transmitter and receiver devices. The carrier frequency of 10.7 MHz is the intermediate frequency of FM radio receivers, meaning that the proposed system can make use of a wide selection of inexpensive, commercial radio frequency devices.     

神经网络集成模型在入侵检测中的应用

入侵检测是网络安全研究中的热点。提出了一种用于入侵检测的神经网络集成模型。该模型采用神经网络集成分类技术,去除训练集中的冗余数据,利用遗传算法优化成员网络的权值,在此基础上训练成员网络,最终通过神经网络对成员网络的输出结果进行融合。理论和实验表明,模型具有较好的检测能力。     

A spark-based big data analysis framework for real-time sentiment prediction on streaming data

There are many data sources that produce large volumes of data. The Big Data nature requires new distributed processing approaches to extract the valuable information. Real-time sentiment analysis is one of the most demanding research areas that requires powerful Big Data analytics tools such as Spark. Prior literature survey work has shown that, though there are many conventional sentiment analysis researches, there are only few works realizing sentiment analysis in real time. One major point that affects the quality of real-time sentiment analysis is the confidence of the generated data. In more clear terms, it is a valuable research question to determine whether the owner that generates sentiment is genuine or not. Since data generated by fake personalities may decrease accuracy of the outcome, a smart/intelligent service that can identify the source of data is one of the key points in the analysis. In this context, we include a fake account detection service to the proposed framework. Both sentiment analysis and fake account detection systems are trained and tested using Naïve Bayes model from Apache Spark's machine learning library. The developed system consists of four integrated software components, ie, (i) machine learning and streaming service for sentiment prediction, (ii) a Twitter streaming service to retrieve tweets, (iii) a Twitter fake account detection service to assess the owner of the retrieved tweet, and (iv) a real-time reporting and dashboard component to visualize the results of sentiment analysis. The sentiment classification performances of the system for offline and real-time modes are 86.77% and 80.93%, respectively.     

Neural network constitutive model for rate-dependent materials

Neural network (NN) constitutive model adjusts itself to describe given stress and strain relationship. It is capable of capturing complex material behavior, using stress and strain sets from experiments. This paper presents a rate-dependent NN constitutive model formulation and its implementation in finite element analysis. The proposed NN model is verified for a standard solid viscoelasticity model. The model is then applied to analysis of time-dependent behavior of concrete. The proposed model has potential of capturing any rate-dependent material models, provided enough data sets are given. The issue of what constitutes a sufficient data set to train a neural network constitutive model must be addressed in future research.     

一类半无限规划问题的神经网络算法

考虑了一类带有不等式和等式混合约束的半无限规划问题。通过运用极大熵方法,将多个约束条件的问题转化为单个约束条件的非线性规划模型,并提出了求解它的一个神经网络模型,严格证明了该模型是Lyapunov稳定的,并且在有限时间内收敛到原问题的一个精确解。数值实验表明,新模型不仅可行而且有效。     

基于实数编码遗传算法的神经网络成本预测模型及其应用

在生产过程中,影响产品成本的因素多而复杂,因素之间相互影响,存在耦合现象,因此准确预测成本是一个重要又难以解决的问题.通过遗传算法(Genetic Algorithm)与误差反向传播(Error Back Propagation)神经网络相结合,提出了用实数编码的自适应变异遗传算法训练神经网络权重的混合算法,避免了传统神经网络易陷入局部极小的缺点.以矩阵形式表示产品成本组成,建立了产品成本组成模型,以此为基础建立了考虑成本因素之间互相影响的神经网络产品成本预测模型,并成功应用于某钢铁企业产品成本的预测,提高了预测精度.     

X射线探测器数据采集与实时传输系统

根据自主设计的X射线探测器读出芯片数据量大、速度快和连续性的特点。设计了基于现场可编程逻辑门阵列(FPGA)、第二代双倍数据率同步动态随机存储器(DDR2)与第三代通用串行总线(USB3. 0)的X射线探测器数据采集与实时传输系统。系统以FPGA和USB 3. 0协议为核心,由FPGA采集X射线探测器读出芯片的24通道数字信号,并送入DDR2存储器中进行高速缓存,高速缓存的数据最终被USB 3. 0接口传输至PC端进行实时处理。该系统数据采集的吞吐率可达2 400 Mbps,USB接口传输速率可达2. 5 Gbps,能够满足X射线探测系统中高速采集与实时传输的要求。     

Neural network and GA approaches for dwelling fire occurrence prediction

This paper describes three approaches for the prediction of dwelling fire occurrences in Derbyshire, a region in the United Kingdom. The system has been designed to calculate the number of fire occurrences for each of the 189 wards in the Derbyshire. In terms of the results from statistical analysis, eight factors are initially selected as the inputs of the neural network. Principal Component Analysis (PCA) is employed for pre-processing the input data set to reduce the number of the inputs. The first three principal components of the available data set are chosen as the inputs, the number of the fires as the output. The first approach is a logistic regression model, which has been widely used in the forest fire prediction. The prediction results of the logistic regression model are not acceptable. The second approach uses a feed-forward neural network to model the relationship between the number of fires and the factors that influence fire occurrence. The model of the neural network gives a prediction with an acceptable accuracy for the fires in dwelling areas. Genetic algorithms (GAs) are the third approach discussed in this study. The first three principle components of the available data set are classified into the different groups according to their number of fires. An iterative GA is proposed and applied to extract features for each data group. Once the features for all the groups have been identified the test data set can be easily clustered into one of the groups based on the group features. The number of fires for the group, which the test data belongs to, is the prediction of the fire occurrence for the test data. The three approaches have been compared. Our results indicate that the neural network based and the GA based approaches perform satisfactorily, with MSEs of 2.375 and 2.875, respectively, but the GA approach is much better understood and more transparent.