outlier resistance

船舶阻力的相关计算涉及多个因素，包括水流速度、船体形状和材料特性。通过合理的模型和公式，可以有效预测船舶在水中行驶时的阻力，从而优化设计，提高效率。

异常（Outlier）指的是数据集中与大部分数据显著偏离的数据点，其偏离程度超出随机因素的范围，可能源于完全不同的生成机制。根据Hawkins的定义，异常是数据中那些使人怀疑其生成方式不同于其他数据的点。根据Weisberg的看法，异常是不符合数据集其他部分统计模型的数据。Samuels认为，异常是与数据集中其余部分显著不同的数据点。Porkess指出，异常是远离数据集中其他数据点的极端值。

6.1 Quantized Conduction Gate Resistance The conduction gate resistance Rg(ON) is selected based on the switching time tsw to achieve the desired switching performance. To calculate the gate resistance, we need to know the power supply voltage VDD (or VBS), the equivalent on-state resistance of the gate driver RDRV(ON), and the switching parameters of the device (Qg, Qgd, and Vgs(th)). The switching time is defined as the time taken to reach the platform voltage, where the total charge Qgd + Qgd is provided to the MOSFET, as shown in Figure 21.The calculation for the gate conduction resistance is as follows:[ Rg(ON) = \text{Function of } VDD, Qg, Qgd, Vgs(th), \text{and } RDRV(ON) ] ### 6.2 Output Voltage SlopeThe conduction gate resistance Rg(ON) also controls the output voltage slope dVOUT/dt. When the output voltage is nonlinear, the large output voltage slope can be approximated as:[ \frac{dVOUT}{dt} = \frac{Ig(avr)}{Cgd(off)} ]Where Cgd(off) represents the Miller capacitance (also defined as Crss in the datasheet). ### 6.3 Quantized Gate Resistance in Off-StateThe quantization of the off-state gate resistance occurs when external actions force rectification when the MOSFET's drain is in the off-state. In this case, the output node's dV/dt induces a parasitic current to flow through Cgd, towards RG(OFF) and RDRV(OFF), as shown in Figure 22.The gate threshold voltage Vgs(th) and the drain-source dV/dt are related in the following equation:[ ISINK \geq \frac{1.5 \times QG}{tSW} ]Finally, the total resistance is determined by:[ RTOTAL = Rg(ON) + RDRV(ON) + \frac{VDD}{Vgs + Ig(avr)} ] This equation links the total resistance with the gate voltage and output voltage slope, ensuring that the switching characteristics meet the specified parameters.

xgboost代码回归matlab在度量学习背景下进行上下文空间离群值检测。此代码由宾夕法尼亚州立大学的Guanguan Zheng编写，并应用于kdd 2017论文：“通过度量学习进行上下文空间离群值检测”，收录于第23届ACM SIGKDD国际会议论文集第2161-2170页。该代码的度量学习部分（位于MLKR1.0文件夹中）由Kilian Q.Weinberger编写，用于内核回归。我们对其进行了修改以增强输入输出功能，并计划更新以兼容Windows和Mac OS系统。建议在anaconda3环境下安装Python 3，并配备Matlab及相关Python包（如xgboost）。

取样是一种普遍有效的近似技术，利用取样技术处理近似聚集查询在决策支持系统和数据挖掘中广泛采用。正确有效地产生近似查询结果并最小化误差是处理近似查询的关键目标。提出了Outlier-DivideConquer算法，结合误差确界，只需单次扫描数据集，特别适用于高方差分布的聚集属性，克服了随机均匀取样的局限性，显著降低近似查询误差，且比同类算法执行效率更高。实验验证显示，该算法与传统均匀取样算法相比具有显著优势。