We introduce a new interpretation and quantitative method for computerized diplopia test. By comparing this new method to the Hess screen test, we validate its applicability among 304 patients with ocular motor nerve ...We introduce a new interpretation and quantitative method for computerized diplopia test. By comparing this new method to the Hess screen test, we validate its applicability among 304 patients with ocular motor nerve palsy. This new method shows great assistant value as the Hess screen test in making accurate diagnosis and quantitative evaluation the severity of diplopia. Furthermore, it is more convenient and suitable for daily clinical use.展开更多
In the existing landslide susceptibility prediction(LSP)models,the influences of random errors in landslide conditioning factors on LSP are not considered,instead the original conditioning factors are directly taken a...In the existing landslide susceptibility prediction(LSP)models,the influences of random errors in landslide conditioning factors on LSP are not considered,instead the original conditioning factors are directly taken as the model inputs,which brings uncertainties to LSP results.This study aims to reveal the influence rules of the different proportional random errors in conditioning factors on the LSP un-certainties,and further explore a method which can effectively reduce the random errors in conditioning factors.The original conditioning factors are firstly used to construct original factors-based LSP models,and then different random errors of 5%,10%,15% and 20%are added to these original factors for con-structing relevant errors-based LSP models.Secondly,low-pass filter-based LSP models are constructed by eliminating the random errors using low-pass filter method.Thirdly,the Ruijin County of China with 370 landslides and 16 conditioning factors are used as study case.Three typical machine learning models,i.e.multilayer perceptron(MLP),support vector machine(SVM)and random forest(RF),are selected as LSP models.Finally,the LSP uncertainties are discussed and results show that:(1)The low-pass filter can effectively reduce the random errors in conditioning factors to decrease the LSP uncertainties.(2)With the proportions of random errors increasing from 5%to 20%,the LSP uncertainty increases continuously.(3)The original factors-based models are feasible for LSP in the absence of more accurate conditioning factors.(4)The influence degrees of two uncertainty issues,machine learning models and different proportions of random errors,on the LSP modeling are large and basically the same.(5)The Shapley values effectively explain the internal mechanism of machine learning model predicting landslide sus-ceptibility.In conclusion,greater proportion of random errors in conditioning factors results in higher LSP uncertainty,and low-pass filter can effectively reduce these random errors.展开更多
This article presents a current EU project on the interpretation of Gastronomic European cultural heritage.Europe’s cultural heritage is highly diverse.This is by no means limited to museums,theatres,or castles;gastr...This article presents a current EU project on the interpretation of Gastronomic European cultural heritage.Europe’s cultural heritage is highly diverse.This is by no means limited to museums,theatres,or castles;gastronomic traditions are also part of Europe’s cultural heritage.As an essential part of the collective European memory,it is essential to preserve this diversity.From an economic perspective,the preservation of cultural heritage is a crucial task for the future.Within the EU,more than 300,000 people work in the cultural heritage sector;in addition,there are about 7.8 million jobs in the EU that correlate indirectly with cultural heritage(e.g.,tourism).How could gastronomic traditions be used to support a sustainable tourisms development?展开更多
以深度神经网络(deep neural network,DNN)为基础构建的自动驾驶软件已成为最常见的自动驾驶软件解决方案.与传统软件一样,DNN也会产生不正确输出或意想不到的行为,基于DNN的自动驾驶软件已经导致多起严重事故,严重威胁生命和财产安全....以深度神经网络(deep neural network,DNN)为基础构建的自动驾驶软件已成为最常见的自动驾驶软件解决方案.与传统软件一样,DNN也会产生不正确输出或意想不到的行为,基于DNN的自动驾驶软件已经导致多起严重事故,严重威胁生命和财产安全.如何有效测试基于DNN的自动驾驶软件已成为亟需解决的问题.由于DNN的行为难以预测和被人类理解,传统的软件测试方法难以适用.现有的自动驾驶软件测试方法通常对原始图片加入像素级的扰动或对图片整体进行修改来生成测试数据,所生成的测试数据通常与现实世界差异较大,所进行扰动的方式也难以被人类理解.为解决上述问题,提出测试数据生成方法IATG(interpretability-analysis-based test data generation),使用DNN的解释方法获取自动驾驶软件所做出决策的视觉解释,选择原始图像中对决策产生重要影响的物体,通过将其替换为语义相同的其他物体来生成测试数据,使生成的测试数据更加接近真实图像,其过程也更易于理解.转向角预测模型是自动驾驶软件决策模块重要组成部分,以此类模型为例进行实验,结果表明解释方法的引入有效增强IATG对转向角预测模型的误导能力.此外,在误导角度相同时IATG所生成测试数据比DeepTest更加接近真实图像;与semSensFuzz相比,IATG具有更高误导能力,且IATG中基于解释分析的重要物体选择技术可有效提高semSensFuzz的误导能力.展开更多
文摘We introduce a new interpretation and quantitative method for computerized diplopia test. By comparing this new method to the Hess screen test, we validate its applicability among 304 patients with ocular motor nerve palsy. This new method shows great assistant value as the Hess screen test in making accurate diagnosis and quantitative evaluation the severity of diplopia. Furthermore, it is more convenient and suitable for daily clinical use.
基金This work is funded by the National Natural Science Foundation of China(Grant Nos.42377164 and 52079062)the National Science Fund for Distinguished Young Scholars of China(Grant No.52222905).
文摘In the existing landslide susceptibility prediction(LSP)models,the influences of random errors in landslide conditioning factors on LSP are not considered,instead the original conditioning factors are directly taken as the model inputs,which brings uncertainties to LSP results.This study aims to reveal the influence rules of the different proportional random errors in conditioning factors on the LSP un-certainties,and further explore a method which can effectively reduce the random errors in conditioning factors.The original conditioning factors are firstly used to construct original factors-based LSP models,and then different random errors of 5%,10%,15% and 20%are added to these original factors for con-structing relevant errors-based LSP models.Secondly,low-pass filter-based LSP models are constructed by eliminating the random errors using low-pass filter method.Thirdly,the Ruijin County of China with 370 landslides and 16 conditioning factors are used as study case.Three typical machine learning models,i.e.multilayer perceptron(MLP),support vector machine(SVM)and random forest(RF),are selected as LSP models.Finally,the LSP uncertainties are discussed and results show that:(1)The low-pass filter can effectively reduce the random errors in conditioning factors to decrease the LSP uncertainties.(2)With the proportions of random errors increasing from 5%to 20%,the LSP uncertainty increases continuously.(3)The original factors-based models are feasible for LSP in the absence of more accurate conditioning factors.(4)The influence degrees of two uncertainty issues,machine learning models and different proportions of random errors,on the LSP modeling are large and basically the same.(5)The Shapley values effectively explain the internal mechanism of machine learning model predicting landslide sus-ceptibility.In conclusion,greater proportion of random errors in conditioning factors results in higher LSP uncertainty,and low-pass filter can effectively reduce these random errors.
基金Acknowledgement:This research was supported in part by a grant from the European Union.
文摘This article presents a current EU project on the interpretation of Gastronomic European cultural heritage.Europe’s cultural heritage is highly diverse.This is by no means limited to museums,theatres,or castles;gastronomic traditions are also part of Europe’s cultural heritage.As an essential part of the collective European memory,it is essential to preserve this diversity.From an economic perspective,the preservation of cultural heritage is a crucial task for the future.Within the EU,more than 300,000 people work in the cultural heritage sector;in addition,there are about 7.8 million jobs in the EU that correlate indirectly with cultural heritage(e.g.,tourism).How could gastronomic traditions be used to support a sustainable tourisms development?
文摘以深度神经网络(deep neural network,DNN)为基础构建的自动驾驶软件已成为最常见的自动驾驶软件解决方案.与传统软件一样,DNN也会产生不正确输出或意想不到的行为,基于DNN的自动驾驶软件已经导致多起严重事故,严重威胁生命和财产安全.如何有效测试基于DNN的自动驾驶软件已成为亟需解决的问题.由于DNN的行为难以预测和被人类理解,传统的软件测试方法难以适用.现有的自动驾驶软件测试方法通常对原始图片加入像素级的扰动或对图片整体进行修改来生成测试数据,所生成的测试数据通常与现实世界差异较大,所进行扰动的方式也难以被人类理解.为解决上述问题,提出测试数据生成方法IATG(interpretability-analysis-based test data generation),使用DNN的解释方法获取自动驾驶软件所做出决策的视觉解释,选择原始图像中对决策产生重要影响的物体,通过将其替换为语义相同的其他物体来生成测试数据,使生成的测试数据更加接近真实图像,其过程也更易于理解.转向角预测模型是自动驾驶软件决策模块重要组成部分,以此类模型为例进行实验,结果表明解释方法的引入有效增强IATG对转向角预测模型的误导能力.此外,在误导角度相同时IATG所生成测试数据比DeepTest更加接近真实图像;与semSensFuzz相比,IATG具有更高误导能力,且IATG中基于解释分析的重要物体选择技术可有效提高semSensFuzz的误导能力.
