Warning: statistical benchmarking is addictive. Kicking the habit in machine learning

Algorithm performance evaluation is so entrenched in the machine learning community that one could call it an addiction. Like most addictions, it is harmful and very difficult to give up. It is harmful because it has serious limitations. Yet, we have great faith in practicing it in a ritualistic manner: we follow a fixed set of rules telling us the measure, the data sets and the statistical test to use. When we read a paper, even as reviewers, we are not sufficiently critical of results that follow these rules. Here, we will debate what are the limitations and how to best address them. This article may not cure the addiction but hopefully it will be a good first step along that road.     

载车SINS/OD杆臂在线补偿算法

在捷联惯性导航系统与里程计构成的车载组合导航系统中,考虑到工程实际,往往将里程计安装在载车前轮,而将惯性测量单元安装在车辆的其他任何位置,比如车的尾部。由于空间布局的不重合将导致里程计和惯性导航系统得到的车速真实信息并不相同,此即车载杆臂效应。为了实现高精度的车载组合导航,必须将此杆臂效应精确补偿。提出首先将载车杆臂列入组合导航系统状态,利用实际跑车机动对其进行精确估计,进而利用估计结果对里程计速度量测进行修正,从而提高整个组合导航系统的精度。跑车试验证明了该方案的有效性。     

弹药步进应力寿命试验数据处理

弹药的步进应力加速寿命试验数据处理的关键是异常数据的处理和弹药长贮寿命分布参数的估计.针对这两个问题,人们提出了许多相关解决办法,但迄今为止,尚没有一种通用的、行之有效的方法.为此,尝试依据贝叶斯理论对试验中可能出现的数据异常进行预处理,并利用鲍维尔法先求取各级加速应力下的样品分布参数的似然估计,再结合加速方程得到常应力下弹药的贮存寿命分布参数.最后,通过实例说明了方法的可行性.     

基于Authentication Test方法的高效安全IKE形式化设计研究

基于Authentication Test方法,围绕高效安全Internet密钥交换（ESIKE）协议的安全目标,提出一种具体地构建唯一满足两个通信实体变换边的形式化协议设计方法,设计出了高效安全的IKE协议;并且基于Strand Space模型和Authentication Test方法,形式化分析ESIKE协议,证明了其所具有的安全特性．该ESIKE协议克服了原有Internet密钥交换（IKE）协议存在的安全缺陷,提供了安全的会话密钥及安全关联（SA）协商,保护了通信端点的身份,并且保证了协议发起者和响应者间的双向认证．同时,ESIKE仅需3条消息及更少的计算量,更加简单、高效．     

石油管材拉伸比对试验及结果分析

为了促进石油管材力学性能试验室检测结果的一致性,国家石油管材质量监督检验中心发起了“石油管材拉伸比对试验”。采用Z比分数法对14家实验室获得的数据进行了分析,通过本次比对试验的实施,发现了检测实验室之间存在的差异。讨论了影响试验结果可靠性的测量因素,试验数据从总体分散性角度看,R_m、R_(t0.5)与A_(50)的离散性依次增大。     

考虑软土流变特性和应力历史的一维固结与渗透试验

利用GDS先进固结试验系统对萧山软土进行了常规分级加载固结与渗透试验、分级加载流变固结与渗透试验、压缩－回弹－再压缩流变固结与渗透试验,绘制了沉降曲线、孔压消散曲线以及孔隙比和渗透系数的变化曲线,通过回归分析得到了萧山软土的压缩指数、回弹指数、渗透指数、次固结系数。试验结果表明：萧山软土具有明显的流变特性,对于正常固结萧山软土,土样底部超静孔压消散大约在加载后5个小时左右完成,而超固结土孔压消散速度比正常固结土快;应力历史和超固结比对次固结系数有显著的影响。     

滩地刚性植物消浪作用特性研究

滩地植物有利于保障沿海地区的水安全,量化成熟期植物消浪作用是提升生态海岸韧性的基础。本文基于一般试验规律和两种波高消减模型,分析提出了新的滩地植物消浪系数计算方法,结合波高条件可计算林带消浪后波高,相关表达式由大量试验数据进行了验证,最后对比了由本文方法与规范方法计算的林带后波高。结果表明,阻尼因数及消浪系数能直观体现植物消浪作用特性,波浪传播方向上植物面积占横截面比例是影响植物消浪机理的关键无量纲参数。植物非淹没时,本文方法得到的林带后波高与规范方法接近;植物淹没时,本文方法能更准确描述植物消浪作用。相较于规范方法,本文方法考虑了植物淹没情况,直观揭示了林带密度、宽度、高度,植物平均直径,水深等参数的影响,适用于更多工况。基于植物消浪作用的模型可快速研判相关参数变化对植物消浪效果的影响,为生态海堤滩地刚性植物规划设计及数模研究提供科学和技术支撑。     

基于真实交通状况的固定式机动车现场测速标准装置

为了满足机动车测速仪现场检测新方法的需要,研制了基于真实交通流量状态下的固定式机动车现场测速标准装置。介绍了该套现场标准装置的设计思路与设计方案,分析了现场标准装置中所用的现场标准测速仪的工作原理,并对其现场速度测量的不确定度进行评估。不确定度评估结果表明:在20～250km/h速度测量范围内,现场标准测速仪的相对扩展不确定度为0.12%(k=3)。通过2款机动车雷达测速仪样品的大样本现场试验数据,验证了机动车测速仪现场检测新方法的可行性,以及该套现场标准装置在真实交通流量状态下的实际测速性能。     

Relationship between NP GCL Internal and HDPE GMX/NP GCL Interface Shear Strengths

An investigation was conducted on the relationship between the internal shear strength of hydrated needle-punched (NP) geosynthetic clay liners (GCLs) and the interface shear strength between hydrated NP GCLs (nonwoven side) and high-density polyethylene (HDPE) textured geomembranes (GMXs). New large-scale direct shear data are presented and compared to previous results obtained using similar materials and procedures. The data indicate that both GCLs and GMX/GCL interfaces display large postpeak strength reduction, even at high normal stress. Peak and large-displacement failure envelopes are nonlinear; except for the GCL internal residual strength envelope, which passes through the origin and has a friction angle of 4.8°. GMX/GCL interfaces can be expected to have lower peak strengths and higher large-displacement strengths than GCL internal shear specimens. However, the failure mode for GMX/GCL specimens can change from interface shear to GCL internal shear as normal stress increases. Design for peak strength conditions should be based on the lowest peak strength interface in a liner system, and design for large displacement conditions should be on the basis of the residual strength of the same interface.     

On the mechanical characterization of materials by Arcan-type specimens

The simple circular notched specimen was originally proposed by Arcan to characterize the elastic properties of fibre-reinforced composites. Unfortunately, its optimized geometry does not allow to measure with reasonable accuracy both the material shear strength and the conditions of failure under a generic biaxial stress state, since the effects of stress concentration on the fillets of the two V-grooves and on the inner circular edges are responsible of premature fractures due to the uniaxial stress states of the notch edges.In a previous numerical study carried out by a parametric two-dimensional finite element model, some of the Authors of this paper found a new optimal geometry of the Arcan specimen able to minimize the notch effect and achieve a uniform pure shear stress field in the gauge cross-section. In the present paper, starting from such a geometry, a new type of Arcan specimen is proposed, having not uniform thickness. An extensive three-dimensional parametric finite element analysis has been done to define its optimal shape. The numerical results show that the new specimen is able to achieve, with a higher probability, material fracture in the minimum cross-section under a pure shear stress distribution which is more uniform than those acting in the Arcan specimen typologies until now proposed.