Steganalysis of neural networks based on parameter statistical bias

Many pretrained deep learning models have been released to help engineers and researchers develop deep learning-based systems or conduct research with minimall effort. Previous work has shown that at secret message can be embedded in neural network parameters without compromising the accuracy of the model. Malicious developers can, therefore, hide malware or other baneful information in pretrained models, causing harm to society. Hence, reliable detection of these vicious pretrained models is urgently needed. We analyze existing approaches for hiding messages and find that they will ineluctably cause biases in the parameter statistics. Therefore, we propose steganalysis methods for steganography on neural network parameters that extract statistics from benign and malicious models and build classifiers based on the extracted statistics. To the best of our knowledge, this is the first study on neural network steganalysis. The experimental results reveal that our proposed algorithm can effectively detect a model with an embedded message. Notably, our detection methods are still valid in cases where the payload of the stego model is low.     

基于动态博弈的补贴模式对机场群航线网络结构的影响研究

为探讨不同补贴模式对机场群航线网络结构的影响，本文构建了关于机场、航空公司和乘客之间的动态博弈模型，通过逆向归纳法得到了纳什均衡条件. 研究结果发现：（1）不同的补贴模式对多机场航线网络结构产生重要影响，低补贴模式促使机场群倾向于形成点对点航线网络结构，反之，在高补贴模式下机场群易于形成枢纽辐射航线网络； （2）在较低的空铁联运总运营成本环境下，补贴竞争使得航空公司倾向于选择系统内的大型机场作为枢纽机场；反之，如果空铁联运总运营成本相对较高，补贴竞争使得航空公司倾向于选择小机场作为枢纽节点； （3）随航班固定成本和空铁联运总运营成本的增大，大型机场作为枢纽节点的概率逐渐减少，而小机场被选为枢纽节点的概率将逐渐增大. 随着机场群内大型机场腹地服务人口比例的增大，航空公司选择点对点航线网络的概率在减小. 研究结论能够为民航管理部门科学制定和应用航线补贴政策提供决策依据.     

Moving beyond Volatility Index (VIX): HARnessing the term structure of implied volatility

This paper considers how information from the implied volatility (IV) term structure can be harnessed to improve stock return volatility forecasting within the state-of-the-art HAR model. Factors are extracted from the IV term structure and included as exogenous variables in the HAR framework. We found that including slope and curvature factors leads to significant forecast improvements over the HAR benchmark at a range of forecast horizons, compared with the standard HAR model and HAR model with VIX as IV information set.     

Theoretical and experimental design of Pt-Co(OH)_2 electrocatalyst for efficient HER performance in alkaline solution

Design efficient electrocatalysts become significant for the hydrogen evolution reaction (HER) in alkaline solution.Here,the HER performance on Pt-Co(OH)2 has been investigated by using DFT calculations.Compared with pure Pt,the Pt-Co(OH)2 shows excellent hydrogen binding energy (-0.29 eV) and hydrogen adsorption free energy (0.02 eV).More strikingly,the water dissociation process on Pt-Co(OH)2 is an exothermic reaction while that of Pt is an endothermic reaction,demonstrating that the Pt-Co(OH)_2 is an excellent catalyst for HER in alkaline medium.Then the two-step electrodepositions was used to fabricate Pt-Co(OH)2 on the carbon paper (CP).The Pt-Co(OH)_2/CP shows the lowest overpotential of only 37.6 mV at a current density of 10 mA cm~(-2) in 0.1M KOH,the smallest Tafel slope of58 mV dec~(-1) and the good stability for 20000s without obvious degradation.     

动车所调车作业计划优化模型与算法

本文在分析动车组一级修作业流程特点的基础上,将动车所调车作业计划优化问题构建为整数规划模型并设计了融合"股道均衡分配规则""股道无效占用时间最小化规则""冲突消解策略"的启发式算法对问题进行求解,以动车所实际案例为背景验证了模型和算法的有效性。结果表明,所提出优化模型和算法是有效的;所提启发式算法相比三类常见的调度方法具有明显的优越性。     

基于BIM技术的室内照明仿真模拟软件计算精度解析

建筑照明专业参与BIM （building information modeling）技术的同平台作业还处于起步阶段。DIALux evo 8.2和ElumTools 2019.07是目前仅有的参与BIM技术的室内照明仿真模拟软件。采用CIE （Commission Internationale de i''Eclairage）发布的CIE 171：2006—test cases to assess the accuracy of lighting computer programs的5.2至5.8测试用例，通过验证这2款软件的计算精度，分析得出软件性能的优缺点。结果表明：在计算点光源直接照度、面光源直接照度以及光通量守恒性能这3个方面，2款软件均具有较高精度；计算大面积漫反射、遇遮挡和不同反射率这3种情况的间接照度时，DIALux evo 8.2计算精度优于ElumTools 2019.07；计算小面积漫反射的间接照度精度时，ElumTools 2019.07计算精度优于DIALux evo 8.2；计算定向透射性能时，这2款软件精度均比较低。进一步分析得出：核心算法是产生误差的主要原因，2款软件信息交换均存在一定缺陷。因此，采用光子映射算法，创建建筑与照明全信息交换接口将成为未来基于BIM技术的室内照明仿真模拟软件的研发方向。     

基于泊位坐标的无人船自主靠泊仿真研究

针对神经网络自主靠泊控制器只能完成特定港口的靠泊,而无法拓展到其他没有训练数据的泊位这一问题,提出一种以泊位顶点为原点,岸线为y轴的坐标系(泊位坐标系)。采用泊位坐标系中相对位置来训练神经网络控制器,在V.Dragon-5000航海模拟器中选取集装箱船“银河号”,在大连港进行靠泊训练后,成功将靠泊控制器拓展到没有训练数据的深圳蛇口港与新加波樟宜港。仿真验证了采用新建立的坐标系可以拓展神经网络自主靠泊控制器适用范围,减小训练数据所需的成本。     

基于神经网络动态优化的再入轨迹设计

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。     

智能下肢助力装置虚拟样机的联合仿真

尝试建立一种全新的智能助力装置虚拟样机模型,旨为提高下肢助力装置设计的效率与可靠性.将CATIA建立的机械模型导入ADAMS建立装置的三维虚拟样机模型,对助力装置的运动学及动力学进行了仿真验证,将虚拟样机模型与MATLAB/simulink结合,对助力装置动力学方程及搭建的控制系统进行联合仿真.根据人体的步态数据,采用PID控制算法对下肢助力系统进行控瓤仿真结果表明,下肢助力控制系统有较好的动态响应特性和轨迹跟踪能力.研究结果能及时发现装置存在的缺点与不足,为助力装置样机研制和电机选型提供了参考,也为实现装置的助力提供了重要的理论依据.