面向网络行为特征分析的网络监测系统设计及实现

网络行为特征分析可为下一代互联网规划设计、建设与管理提供科学依据,对网络行为的监测极为重要.根据网络行为分析的要求,构建了IP网络行为评价指标体系,详细介绍了面向网络行为特征分析的网络监测系统的设计与实现,对系统设计的关键问题做了分析,并搭建实验环境时实现的系统进行了测试.实验结果表明:该系统满足网络行为实时监测各项要求,能为网络的运营管理提供决策依据.     

基于改进YOWO算法的学生课堂行为识别

当前,大部分的学生课堂行为识别工作主要基于单帧图像进行,忽略了行为的连贯性,因此不能充分利用视频信息来对学生的课堂行为进行准确刻画.所以,本文提出一种改进的YOWO算法模型,有效利用视频信息对学生课堂行为进行识别.首先,本文采集某高校真实课堂教学中的授课录像,制作出包含5类学生课堂行为的AVA格式视频数据集;其次,采用时移模块TSM (temporal shift module),用来增强模型获取时间上下文信息的能力;最后,采用非局部操作模块non-local来提高模型提取关键位置信息的能力.实验结果表明,通过对YOWO模型的优化,使得网络的识别性能更佳.在学生课堂行为数据集上,改进后的算法的mAP值为95.7%,相较于原YOWO算法在mAP值上提高了4.6%;模型参数量为81.97×10⁶,计算量为22.6 GFLOPs,参数量和计算量分别降低32.3%和9.6%;检测速度为24.03 f/s,提升了约3 f/s.     

Macro- and Nanoscale Effect of Ethanol on Bovine Serum Albumin Gelation and Naproxen Release

We report extended ethanol-induced gelation procedures of bovine serum albumin (BSA) at 37 °C and investigate the release behavior of a spin-labeled naproxen derivative (SL-NPX) from these hydrogels. The macroscopic mechanical properties of these gels during formation were studied using rheology, while a nanoscopic, more molecular view was obtained by analyzing the secondary structure of the protein during gelation via infrared (ATR-IR) spectroscopy. To evaluate the potential use of BSA hydrogels in controlled drug delivery, SL-NPX-BSA interaction was investigated in detail by continuous-wave electron paramagnetic resonance (CW EPR) spectroscopy, which provides information on the interaction of the small drug molecules and the hydrogel. In addition to CW EPR spectroscopy, dynamic light scattering (DLS), which provides insight into the size and nature of released components, was applied to characterize the combined influence of incubation time, ethanol, SL-drug, and BSA concentration on release behavior. It was found that the alteration of initial drug loading percentage, hydrogel incubation time as well as BSA and alcohol concentrations affect and thus tune the release rate of SL-NPX from BSA hydrogels. These results lead to the conclusion that BSA hydrogels as controlled release systems offer a remarkable fine-tuning capability for pharmaceutical applications due to the variety of gelation parameters.     

Assessing the Morphological and Behavioral Toxicity of Catechol Using Larval Zebrafish

Catechol is a ubiquitous chemical used in the manufacturing of fragrances, pharmaceuticals and flavorants. Environmental exposure occurs in a variety of ways through industrial processes, during pyrolysis and in effluent, yet despite its prevalence, there is limited information regarding its toxicity. While the genotoxicity and gastric carcinogenicity of catechol have been described in depth, toxicological studies have potentially overlooked a number of other effects relevant to humans. Here, we have made use of a general and behavioral larval zebrafish toxicity assay to describe previously unknown catechol-based toxicological phenomena. Behavioral testing revealed catechol-induced hypoactivity at concentrations an order of magnitude lower than observable endpoints. Catechol exposure also resulted in punctate melanocytes with concomitant decreases in the expression of pigment production and regulation markers mitfa, mc1r and tyr. Because catechol is converted into a number of toxic metabolites by tyrosinase, an enzyme found almost exclusively in melanocytes, an evaluation of the effects of catechol on these cells is critical to evaluating the safety of this chemical. This work provides insights into the toxic nature of catechol and highlights the benefits of the zebrafish larval testing platform in being able to dissect multiple aspects of toxicity with one model.     

A Systematic Study of Restorative Crown-Materials Combinations for Dental Implants: Characterization of Mechanical Properties under Dynamic Loads

This study aimed to find the optimum mechanical characteristics of the restorative materials for the manufacture of implant crowns subjected to impact loading when different combinations of materials are used for the inner and outer crown. Several combinations of external–internal crown restorative materials were analyzed. The dynamic stresses at eight different zones of a dental implant subjected to an impact load and the influence of several mechanical properties, such as the Young’s modulus, Poisson’s ratio, density, and initial velocity, were analyzed and compared. A detailed 3D model was created, including the crown, the retention screw, the implant, and a mandible section. The model was then built by importing the 3D geometries from CAD software. The whole 3D model was carefully created in order to guarantee a finite element mesh that produced results adjusted to physical reality. Then, we conducted a numerical simulation using the finite element method (FEM). The results of the FEM analysis allowed for evaluating the effect that different combinations of restorative materials and mechanical properties had on the stress distribution in various regions of the implant. The choice of restorative material is a factor to be considered in order to preserve the integrity of osseointegration. Restorative materials transfer more or less stress to the dental implant and surrounding bone, depending on their stiffness. Therefore, an inadequate Young’s modulus of the rehabilitation material can affect the survival of the implant over time. Eight interactive graphics were provided on a web-based surface platform to help clinical dentists, researchers, and manufacturers to select the best restorative materials combination for the crown.     

机器人足球仿真教练模型研究

RoboCup仿真比赛已经在国际上开展几年了,近来在国内也得到了迅速的发展,它通过提供了一个标准问题来促进人工智能和智能机器人的研究。文中以RoboCup仿真比赛为背景,首先讨论了教练模型,包括教练的角色意义和相应的权限,而后针对教练设计中涵盖的几个问题,如对手建模、规划识别、离线训练,分别给出了较好的解决方案。最后从在线教练工作过程的角度总结了有关教练技术的具体应用,并对教练的设计前景提出了一些看法。对于建立和设计仿真教练,进行多智能体系统的合作协调研究具有重要意义。     

基于EBP的宏观网络流量异常行为检测

针对以往Hurst指数估算方法在求解精度和实时性上的不足,提出将EBP引入到网络流量白相似特性分析中,对比实验表明EBP对Hurst指数的估算更精确、实时性更高。利用EBP的这一优势将其运用到宏观网络行为的在线实时分析和异常行为的检测中,对林肯实验室宏观网络行为数据的分析表明,正常行为和异常行为的Hurst分布曲线差异明显。与传统匹配方法相比,基于EBP的异常行为检测方法检测效率更高。     

Numerical investigation of fracture behavior of tunnel by excavation loading

A rock failure process analysis model, RFPA2D code, a two-dimensional numerical code, were proposed. The code not only satisfied the global equilibrium, strain consistent and nonlinear constitutive relationship of rock and soil materials but also took into account the heterogeneous characteristics of rock materials at macroscopic and microscopically level. The failure behavior of tunnel could be simulated by this numerical model. The model could realistically simulate the fracture behavior of tunnel by excavation loading, strength limits, and post peak response for both tension and compression. As the proposed method was used to conduct the stability analysis of tunnel, the safety factor of tunnel was defined as the ratio of actual shear strength parameter to critical failure shear strength parameter. Not only the safety factor of tunnel with specific physics meaning can be obtained, but also the overall failure process and the location of failure surface may also be determined at the same time.     

基于模糊逻辑的人为差错概率量化方法

为满足人机系统概率风险评估的需要,提出一种人为差错概率量化方法。分析技能、规则和知识为基础 (skill, rule and knowledge-based,SRK)框架和行为模式的确定方法Hanaman 决策树法,指出在确定行为模式的过程 中考虑行为模式影响因素的不确定性是必要的;使用模糊逻辑方法处理行为模式各个影响因素的不确定性,根据 Hanaman 决策树构建模糊推理规则,利用系统人为行为可靠性程序(systematic human action reliability procedure, SHARP)方法所提供的人为差错概率区间确定人为差错概率的隶属度函数。结果表明：该方法考虑了任务场景的不确 定性,可以得到人为差错概率的精确值,满足人机系统概率风险评估的需要。     

动态特征选择算法对恶意行为检测的优化研究

针对互联网中存在的恶意行为,特别是社交网络应用中的在线恶意行为,通常使用基于用户多维特征的聚类分析算法进行检测。提出一种动态特征选择算法(DFSA),使用具有特征加权熵的模糊C均值目标函数,首先为参数构建一个学习模式,自动计算每个特征权重,并剔除权重小于阈值的特征,动态选择重要的特征,迭代地更新隶属函数、簇中心和特征权重直到最优化为止,最后识别出具有高精度的恶意用户行为簇。仿真结果表明,对比SDAFS算法、ELAFC算法和NADMB算法,DFSA算法在Rand指数、Jaccard指数和归一化互信息量3个主要性能指标上均有改善。