基于Class Balanced Loss修正交叉熵的非均衡样本信用风险评价模型

针对传统信用风险预测模型存在对非违约样本识别过度、对违约样本识别不足的问题,将平衡损失Class Balanced Loss函数引入信用风险评价,构建Class Balanced Loss修正交叉熵的非均衡样本信用风险评价模型。利用所建模型与交叉熵神经网络、支持向量机、决策树、随机森林和K最近邻5种分类模型进行对比,验证BPNN-CBCE对中国某金融机构1 534笔农户贷款数据信用风险预测的有效性;在此基础上,利用UCI公开的德国信贷数据验证BPNN-CBCE模型的稳健性。研究表明:对于农户数据,BPNN-CBCE模型在AUC、违约召回率Default recall方面普遍优于BPNN-CE、SVM、DT、RF和KNN模型,其中,BPNN-CBCE的Default recall相比5种对比模型提升了41.3个百分点,AUC相比5种对比模型提升了15.6个百分点;对于德国数据集,BPNN-CBCE评级模型在AUC、违约召回率Default recall方面也均优于5种对比模型。因此,BPNN-CBCE信用评价模型对农户不均衡信贷数据中的违约样本具有较好的识别能力,可有效降低金融机构客户误判带来的损失。创新与特色:①利用Class Balanced Loss中的平衡因子ω,增大违约样本在目标损失中的权重、降低非违约样本在目标损失中的权重,客观调节正负样本损失在目标损失中权重,弥补交叉熵函数无法调节两类样本损失权重的缺陷,克服由样本不均衡带来的评价模型对非违约样本识别过度、对违约样本识别不足。②通过考虑数据重叠,利用随机覆盖方法,分别对贷款数据中违约、非违约样本进行不放回采样,以对全样本空间X_(违约)、X_(非违约)进行不重叠覆盖,计算两类贷款客户的有效样本数量。既反映由于真实数据之间的内在相似性,随着样本数量的增加,新添加样本很可能是现有样本近似重复的客观事实,也保证基于有效样本对两类样本损失进行重新加权的客观性。将图像识别领域中的Class Balanced Loss函数引入信用评价领域,既拓展了Class Balanced Loss的使用边界,也为解决不均衡样本的信用风险评价提供了新的研究思路。

Estimation under a Finite Mixture of Exponentiated Exponential Components Model and Balanced Square Error Loss

By exponentiating each of the components of a finite mixture of two exponential components model by a positive parameter, several shapes of hazard rate functions are obtained. Maximum likelihood and Bayes methods, based on square error loss function and objective prior, are used to obtain estimators based on balanced square error loss function for the parameters, survival and hazard rate functions of a mixture of two exponentiated exponential components model. Approximate interval estimators of the parameters of the model are obtained.

Strong Interaction Effect on Jet Energy Loss with Detailed Balance

The strong force effect on gluon distribution of quark-gluon plasma and its influence on jet energy loss with detailed balance are studied. We solve the possibility equation and obtain the value of non-extensive parameter q. In the presence of strong interaction, more gluons stay at low-energy state than the free gluon case. The strong interaction effect is found to be important for jet energy loss with detailed balance at intermediate jet energy. The energy gain via absorption increases with the strong interaction. This will affect the nuclear modification factor RAA and the parameter of q at intermediate jet energy.

Balance Analysis of Profit and Loss on Exploiting Coal-Bed Gas in China

Models about four aspects according to the balance principle and practice in China were established, which involve the minimum production scale, alert production scale, safe running scale, and the goal production scale for specified profit level. It provides an effective quantitative analyzing method for the investors of coal bed gas exploitation project.

Study on regulation of potassium loss and its balance characteristics from typical paddy soils in Dongting Lake area

Potassium loss due to runoff and leaching, crop uptake, and K-fertilizer input, irrigation and wet deposition was studied, subsequently potassium balance characteristics was analyzed using lysimeter modeled two typical paddy soils from double rice field in Dongting Lake area. K losses due to runoff and leaching in Alluvial Sandy Loamy Paddy Soil （ASL） with lower cation exchange content （CEC） and clayey granule content were increased significantly compared with those in Reddish Yellow Loamy Paddy Soil （RYL） with higher CEC and clayey granule content. Compared with RYL, lower economic and biological yield due to lower capacity of nutrient and water keeping in ASL resulted in lower K uptake. So potassium deficit in double rice field of RYL was more serious than that of ASL because main K proportion was accounted by crop uptake. Available K in soil of surface layer （0-15 cm） was decreased significantly due to long term nutrient management of K deficit in double rice field. The straw return paddy field was suggested as an effective way to increase soil K and balance K deficit in double rice field because large amount of potassium was enriched in rice straw at harvest.

Effects of three types of soil amendments on yield and soil nitrogen balance of maize–wheat rotation system in the Hetao Irrigation Area, China

Excessive fertilization combined with unreasonable irrigation in farmland of the Hetao Irrigation Area(HIR), China, has resulted in a large amount of nitrogen(N) losses and agricultural non-point source pollution.Application of soil amendments has become one of the important strategies for reducing N losses of farmland.However, there is still no systematic study on the effects of various soil amendments on N losses in the HIR.In this study, three types of soil amendments(biochar, bentonite and polyacrylamide) were applied in a maize–wheat rotation system in the HIR during 2015–2017.Yields of maize and wheat, soil NH3 volatilization, N2O emission and NO3– leaching were determined and soil N balance was estimated.The results showed that applications of biochar, bentonite and polyacrylamide significantly increased yields of maize by 9.2%, 14.3% and 13.3%, respectively, and wheat by 9.2%, 16.6% and 12.3%, respectively, compared with the control(fertilization alone).Applications of biochar, bentonite and polyacrylamide significantly reduced soil N leaching by 23.1%, 35.5% and 27.1%, soil NH3-N volatilization by 34.8%, 52.7% and 37.8%, and soil N surplus by 23.9%, 37.4% and 30.6%, respectively.Applications of bentonite and polyacrylamide significantly reduced N2O-N emissions from soil by 37.3% and 35.8%, respectively, compared with the control.Compared with application of biochar, applications of bentonite and polyacrylamide increased yields of maize and wheat by 5.1% and 3.5%, respectively.Our results suggest that soil amendments(bentonite and polyacrylamide) can play important roles in reducing N losses and increasing yield for the maize–wheat rotation system in the HIR, China.

Surface energy balance on the Antarctic plateau as measured with an automatic weather station during 2014

AWS data during 2014 collected at PANDA-N station, on the East Antarctic Plateau, are analysed. Net Short Wave Radiation(QSWR), net Long Wave Radiation(QLWR), sensible(QH), latent(QL) and subsurface or ground(QG) heat fluxes are computed. Annual averages for QSWR, QLWR, QH, QL and QG of 19.65,-49.16, 26.40,-0.77 and 3.86 W·m-2 were derived based on an albedo value of 0.8. QSWR and QH are the major sources of heat gain to the surface and QLWR is the major component of heat loss from the surface. An i terative method is used to estimate surface temperature in this paper;surface temperature of snow/ice is gradually increased or decreased, thereby changing longwave radiation, sensible, latent and subsurface heat fluxes, so that the net energy balance becomes zero. Mass loss due to sublimation at PANDA-N station for 2014 is estimated to be 12.18 mm w.e.·a-1;and mass gain due to water vapour deposition is estimated to be 3.58 mm w.e.·a-1. Thus the net mass loss due to sublimation/deposition is 8.6 mm w.e.·a-1. This study computes surface energy fluxes using a model, instead of direct measurements. Also there are missing data especially for wind speed, though 2 m air temperature data is almost continuously available throughout the year. The uncertainties of albedo, wind speed and turbulent fluxes cause the most probable error in monthly values of QLWR, QH, QL, QG and surface temperature of about ±4%, ±20%, ±50%, ±11% and ±0.74 K respectively.

Pediatric obesity prevention:From na?ve examination of energy imbalance towards strategies that influence the competition for nutrient resources among tissues

Current pediatric obesity interventions have collectively yielded relatively unsuccessful results. In this Field of Vision, we present plausible physiologic underpinnings fostering ineffectiveness of conventional strategies grounded in requisite induction of negative energy imbalance. Moreover, such recommendations exacerbate the underlying metabolic dysfunction by further limiting metabolic fuel availability, lowering energy expenditure, and increasing hunger(recapitulating the starvation response amid apparent nutritional adequacy) which precede and promote obesity during growth and development. The qualitative aspects of musculoskeletal system(i.e., endocrine response, muscle functional capacity) are likely to improve metabolic function and increase nutrient delivery and utilization. An intricate and complex system including multiple feedback mechanisms operates to homeostatically regulate energy balance and support optimal body composition trajectories and metabolic health, during growth and development. Thus, ignoring the interdependencies of regulatory growth processes initiates a nuanced understanding of energy regulation and thus misguided attempts at preventive strategies. Importantly, these gains are not dependent upon weight-loss, rather we suggest can be achieved through resistance training. Collectively, optimizing musculoskeletal health via resistance training elicits augmentation of competitive capacity across systems. Further, substantial gains can be achieved in skeletal muscle mass, strength, and functional capacity through resistance training in a relatively short period of time.

Unbalance Level Regulating Algorithm in Power Distribution Networks

The paper dwells on the unified power quality indexes characterizing the phenomenon of voltage unbalance in three-phase systems. Voltage unbalance is one of the commonest occurrences in the town mains of 0.38 kV voltage. The phenomenon describes as inequality of vector magnitude of phase voltage and shearing angle between them. Causes and consequences of the voltage unbalance in distribution networks have been considered. The algorithm, which allows switching one-phase load, has been developed as one of the methods of reducing the unbalance level. The algorithm is written in the function block diagram programming language. For determining the duration and magnitude of the unbalance level it is proposed to introduce the forecasting algorithm. The necessary data for forecasting are accumulated in the course of the algorithm based on the Function Block Diagram. The algorithm example is given for transforming substation of the urban electrical power supply system. The results of the economic efficiency assessment of the algorithm implementation are shown in conclusion. The use of automatic switching of the one-phase load for explored substation allows reducing energy losses (active electric energy by 7.63%;reactive energy by 8.37%). It also allows improving supply quality to a consumer. For explored substation the average zero-sequence unbalance factor has dropped from 3.59% to 2.13%, and the negative-sequence unbalance factor has dropped from 0.61% to 0.36%.

Intelligent identification method and application of seismic faults based on a balanced classification network

This study combined fault identification with a deep learning algorithm and applied a convolutional neural network(CNN)design based on an improved balanced crossentropy(BCE)loss function to address the low accuracy in the intelligent identification of seismic faults and the slow training speed of convolutional neural networks caused by unbalanced training sample sets.The network structure and optimal hyperparameters were determined by extracting feature maps layer by layer and by analyzing the results of seismic feature extraction.The BCE loss function was used to add the parameter which is the ratio of nonfaults to the total sample sets,thereby changing the loss function to find the reference of the minimum weight parameter and adjusting the ratio of fault to nonfault data.The method overcame the unbalanced number of sample sets and improved the iteration speed.After a brief training,the accuracy could reach more than 95%,and gradient descent was evident.The proposed method was applied to fault identification in an oilfield area.The trained model can predict faults clearly,and the prediction results are basically consistent with an actual case,verifying the effectiveness and adaptability of the method.

Peripheral vestibular system: Age-related vestibular loss and associated deficits

Given the interdependence of multiple factors in age-related vestibular loss (e.g., balance, vision,cognition), it is important to examine the individual contributions of these factors with ARVL. While therelationship between the vestibular and visual systems has been well studied (Bronstein et al., 2015),little is known about the association of the peripheral vestibular system with neurodegenerative disorders (Cronin et al., 2017). Further, emerging research developments implicate the vestibular system asan opportunity for examining brain function beyond balance, and into other areas, such as cognition andpsychological functioning. Additionally, the bidirectional impact of psychological functioning is understudied in ARVL. Recognition of ARVL as part of a multifaceted aging process will help guide thedevelopment of integrated interventions for patients who remain at risk for decline. In this review, wewill discuss a wide variety of characteristics of the peripheral vestibular system and ARVL, how it relatesto neurodegenerative diseases, and correlations between ARVL and balance, vision, cognitive, and psychological dysfunction. We also discuss clinical implications as well as future directions for research, withan emphasis on improving care for patients with ARVL.

Fast assessment of earthquake loss and its application to the 2008 M_S8.0 Wenchuan earthquake

The paper introduces firstly the seismic loss assessment method based on c indicators and new vulnerability models determined by the data from the on-site damage and loss survey to earthquakes occurred in China during the last two decades. The fast assessment for the 2008 Wenchuan earthquake with Ms8.0 is given based on an empirical intensity attenuation relationship. Compared with the assessment based on the practical seismic intensity map of the event according to the on-site investigation, the result demonstrates the usability of the seismic vulnerability models introduced in the paper. In addition, it is indicated that the main uncertainty of losses in the fast loss assessment comes from the uncertainty of the estimation of seismic ground motion.

Simulation of Irrigation Water Loss Based on VSMB Model

The low degree of development and utilization as well as the contradiction between supply and demand of water resources in Huangshui River basin are the main restricting factors of the local agricultural development. The study on the simulation of irrigation water loss based on the VSMB model has very important significance to strengthening regional water management and improving water resource utilization efficiency. Five groundwater wells were set up to carry out the farmland irrigation water infiltration and the experimental study on groundwater dynamic effect. Two soil moisture monitoring sites were set up in two typical plots of Daxia and Guanting irrigation area at the same time and TDR300 was used to monitor four kinds of deep soil moisture( 10 cm,30 cm,50 cm and 70 cm). On this basis,the VSMB model was used to study the irrigation water loss in the irrigation area of Yellow River valley of Qinghai Province,including soil moisture content,the actual evapotranspiration,infiltration,runoff,groundwater buried depth and so on. The results showed that the water consumption caused by soil evaporation and crop transpiration accounted for 46. 4% and 24. 1% of the total precipitation plus irrigation,respectively,and the leakage accounted for 30. 3% and 60. 6% of the total precipitation plus irrigation,respectively,from March 1,2013 to April 30,and from August 1 to September 30. The actual evaporation of the GT- TR1 and GT- TR2 sites in the whole year of 2013 was 632. 6 mm and 646. 9 mm,respectively,and the leakage accounted for 2. 6% and 1. 2% of the total precipitation plus irrigation,respectively. RMSE of the simulation results of the groundwater depth in Daxia irrigation area during the two periods was 92. 3 mm and 27. 7 mm,respectively. And RMSE of the simulation results of the water content of soil profile in the two monitoring sites of Guanting irrigation area was 2. 04% and 5. 81%,respectively,indicating that the simulation results were reliable.

Optimal Placement of Reclosers to Minimize Power Losses during Non-Load Bearing Disturbances in a Power Distribution Using Firefly Algorithm

The research reported in this paper focuses on non-technical power loss reduction for power distribution systems. Such reduction of costs of energy not served (ENS.COST), is intelligently evaluated and optimized using a firefly algorithm, from where savings of 43.3% on energy not served are achieved.

Effects of Water Weight-Loss Walking Training on Lower Limb Motor Function and Gait in Stroke Patients

Background: Water weight-loss walking training is an emerging physical therapy technique, which provides new ideas for improving the motor function of stroke patients and improving the quality of life of patients. However, the rehabilitation effect of water weight-loss training in stroke patients is currently unclear. Objective: To analyze the effect of water weight loss walking training in stroke patients. Methods: A total of 180 stroke patients admitted to our hospital from January 2019 to December 2021 were selected and randomly divided into two groups. The control group received routine walking training, and the research group performed weight loss walking training in water on this basis. The lower limb motor function, muscle tone grade, daily living ability, gait and balance ability were compared between the two groups before and after treatment. Results: Compared with the control group, the FMA-LE score (Fugl-Meyer motor assessment of Lower Extremity), MBI score (Modified Barthel Index) and BBS score (berg balance scale) of the study group were higher after treatment, and the muscle tone was lower (P Conclusion: Water weight loss walking training can enhance patients’ muscle tension, correct patients’ abnormal gait, improve patients’ balance and walking ability, and contribute to patients’ motor function recovery and self-care ability improvement.

适用于中点箝位型三电平逆变器的多目标调制策略

为同时实现中点箝位型三电平逆变器的中点电压平衡和共模电压降低,分别分析了基于载波脉宽调制(CBPWM)和基于虚拟空间矢量脉宽调制(VSVPWM)的中点电压平衡和共模电压降低方法。基于CBPWM提出一种可以降低共模电压的载波脉宽调制(RCMV_CBPWM)策略,可将大多数工作区域的共模电压限制在直流侧电压的1/6以内。为了实现上述2个目标,同时不过分增加开关损耗,在RCMV_CBPWM基础上提出了一种多目标调制(MOPWM)策略,在全调制度和全负载功率因数范围内,共模电压可限制在直流侧电压的1/6以内并实现中点电压平衡,且其开关损耗低于VSVPWM。比较了MOPWM、RCMV_CBPWM和VSVPWM在中点电压控制、共模电压降低和开关损耗降低方面的性能。实验结果验证了MOPWM的可行性和优越性。

作业疗法联合减重平板步行训练对脑卒中偏瘫患者平衡与功能性步行能力的影响

目的:探讨作业疗法联合减重平板步行训练对脑卒中偏瘫患者平衡与功能性步行能力的影响。方法:选取2019年7月—2021年6月河南科技大学第一附属医院新区医院康复医学科收治的90例脑卒中偏瘫患者作为研究对象,采用随机数表法分为试验组(常规康复护理)与对照组(作业疗法联合联合减重平板步行训练),每组各45例。比较两组患者干预前及干预3个月的平衡功能与功能性步行能力。结果:干预3个月,试验组患者Berg平衡量表(BBS)、功能性步行量表(FAC)评分显著高于对照组,差异有统计学意义(t=23.279、5.555,P<0.05);试验组患者脑卒中专用生活质量量表(SS-QOL)各项评分显著高于对照组,差异有统计学意义(t=8.424、8.637、26.306、5.162、7.579、15.302、8.918、8.080、20.745、10.862、1.397、6.806,P<0.05)。结论:作业疗法联合减重平板步行训练可有效提高脑卒中偏瘫患者平衡功能及功能性步行能力,有利于改善患者的生活质量。

多执行器载荷差异储能均衡系统特性

针对阀控多执行器复合作业系统节流损失大、动势能浪费严重等问题,提出一种多执行器载荷差异储能均衡原理。首先对多执行器系统功率分配特性进行分析,明确了载荷差异产生节流损失的原因是动力源压力与最大负载相匹配,导致其他执行器控制阀产生过大的压力损失。然后建立了载荷储能均衡液压挖掘机联合仿真模型。并设计电液储能单元控制各执行器进油腔的压力相等,从而消除执行器载荷差异造成的节流损失。在动臂、铲斗同时动作时,与流量匹配系统相比,所提系统节流损失降低达75%,系统效率提升39%;同时,电液储能单元实现了传统压力补偿器的压差调控功能,显著提升了系统运行平稳性。

一种三相双输出有源中点钳位型三电平变换器

该文基于有源中点钳位型三电平变换器(ANPC-TLC)结构,通过每相增加两个开关管形成一种具有两组三相交流电压输出端口的双输出有源中点钳位型三电平变换器(DOANPC-TLC)拓扑。该文详细介绍DO-ANPC-TLC的拓扑结构,对其工作原理进行分析,研究其不同开关状态下的电流路径、输出相电压电平和各开关的电压应力。针对DO-ANPC-TLC结构特点及工作原理,应用一种双输出虚拟空间矢量脉宽调制(DO-VSVPWM)方法,能够独立控制两组交流输出,并保持直流链路电容电压平衡。为了进一步简化控制过程,研究一种基于载波的简化双输出虚拟空间矢量脉宽调制(CB-VSVPWM)策略,在器件损耗和总谐波畸变率(THD)方面,对所提出DO-ANPC-TLC的性能进行分析。实验结果验证了所提拓扑的可行性和调制策略的有效性。

基于特征约束和自适应损失平衡的机器遗忘方法

随着数字化进程的加速推进,数据要素已成为现代社会运转的核心驱动力.由于深度学习模型训练需要大量数据作为输入,其数据隐私保护问题也愈发重要.机器遗忘技术使模型能够删除特定数据的影响,同时保持对剩余数据的泛化性能,为深度学习模型中的数据要素安全保护提供了有效的解决方案.现有的机器遗忘方法主要分为精确遗忘和近似遗忘2类,但前者需要干预模型原始训练流程,后者则在遗忘效果和模型泛化能力之间难以找到平衡点.为此,提出了一种基于特征约束和自适应损失平衡的近似遗忘框架.首先,对于“遗忘”过程,使用同样未经过遗忘样本训练的随机模型作为教师来引导遗忘模型的特征输出,实现模型对数据要素在特征层面的遗忘.然后,采用少量剩余数据进行微调训练,来“恢复”模型对于其他数据的泛化性能.将上述机器遗忘框架视为一个多任务优化问题,在“遗忘”和“恢复”2个任务中引入自适应损失平衡,实现任务的稳步训练.以卷积神经网络模型为例,在3个公开数据集上对比了UNSIR等多种基线方法,实验结果表明,该方法构建的遗忘模型不仅保障了机器遗忘效果,在剩余数据的准确率、时间开销、预测结果分布等指标上优于同类方法,更加有效地保护了模型的泛化性能.