![](https://img-blog.csdnimg.cn/20201014180756724.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
论文解读
文章平均质量分 94
年糕糕糕
ML测试
展开
-
[论文解读]Deep Active Learning for Efficient Training of a LiDAR 3D Object Detector
Deep Active Learning for Efficient Training of a LiDAR 3D Object Detector文章目录Deep Active Learning for Efficient Training of a LiDAR 3D Object Detector简介摘要初步相关工作基于激光雷达测点的自动驾驶目标检测深度主动学习方法论Process3D物体探测器不确定度估计和查询功能实验结果实验设计“完美”RGB图像探测器的评价利用预先训练的RGB图像探测器进行评估讨论原创 2021-09-24 14:56:55 · 527 阅读 · 0 评论 -
[论文解读]Active Learning for Deep Object Detection via Probabilistic Modeling
Active Learning for Deep Object Detection via Probabilistic Modeling文章目录Active Learning for Deep Object Detection via Probabilistic Modeling简介摘要初步相关工作主动学习在目标检测中的应用基于概率建模的目标检测提高参数效率计分函数实验基于概率建模的目标检测主动学习评价可扩展性和数据集可传输性结论结论简介论文标题Active Learning for Deep O原创 2021-09-24 14:52:37 · 1700 阅读 · 1 评论 -
[论文解读]Deep active learning for object detection
Deep active learning for object detection文章目录Deep active learning for object detection简介摘要初步以前的工作Method黑箱主动学习3.1.1 Minmax (mm)3.1.2 Maximum Entropy (ent)3.1.3 Sum Entropy (ent-sum)白盒主动学习勘探/开发框架结果PASCAL VOC不同口味白盒的比较与黑箱方法的比较勘探/开发方法比较使用35%图像的不同方法的mapKitti模型分原创 2021-09-22 22:45:58 · 947 阅读 · 0 评论 -
[论文解读] Metamorphic Object Insertion for Testing Object Detection Systems
Metamorphic Object Insertion for Testing Object Detection Systems文章目录Metamorphic Object Insertion for Testing Object Detection Systems简介摘要初步目标检测技术背景方法概述Equality CriteriaOD故障案例分析适用范围设计对象提取对象细化和选择对象插入实现实验评估概述对象位置调整的效率合成图像的自然度使用错误触发图像进行再培训研讨与未来工作Related Work原创 2021-09-22 13:55:36 · 776 阅读 · 0 评论 -
[论文解读]A Quantitative Analysis Framework for Recurrent Neural Network
A Quantitative Analysis Framework for Recurrent Neural Network文章目录A Quantitative Analysis Framework for Recurrent Neural Network简介摘要动机THE DeepStellar FRAMEWORK抽象模型构建应用实施和功能评估相关工作总结与下一步工作简介论文标题A Quantitative Analysis Framework for Recurrent Neural Netw原创 2020-10-22 21:27:06 · 554 阅读 · 0 评论 -
[论文解读]A Survey of Methods for Explaining Black Box Models
A Survey of Methods for Explaining Black Box Models文章目录A Survey of Methods for Explaining Black Box Models简介摘要可解释、可解释和可理解的模型可解释性的维度对可解释模型的渴求打开黑匣子问题问题与基于解释器的分类解决模型解释问题解决结果解释问题基于显著性掩码的深度神经网络解释解决模型检验问题通过敏感性分析进行检查通过部分依赖进行检查通过激活最大化进行检查通过树可视化进行检查解决透明盒的设计问题通过规则提原创 2020-10-21 17:11:55 · 1402 阅读 · 1 评论 -
[论文解读]Towards Improved Testing For Deep Learning
Towards Improved Testing For Deep Learning文章目录Towards Improved Testing For Deep Learning简介摘要初步背景先验文学改善的机会为什么我们需要更好的覆盖标准?为什么我们需要更好的测试输入生成?为什么我们需要一个更好的oracle呢?初步探讨和结果结论简介论文标题Towards Improved Testing For Deep Learning走向深度学习的改进测试2019提出了一种组合测试的覆盖方法,以三元原创 2020-10-12 14:27:03 · 710 阅读 · 6 评论 -
[论文解读]Adaptive Random Testing: The ART of test case diversity
Adaptive Random Testing: The ART of test case diversity文章目录Adaptive Random Testing: The ART of test case diversity简介摘要介绍关于“随机性”的一点注记故障模式自适应随机测试理论极限数字程序以外的ART进一步的影响自适应随机序列基于故障的测试一种软件测试理论测试用例多样性的作用结论简介论文标题Adaptive Random Testing: The ART of test case di原创 2020-09-17 21:28:32 · 1918 阅读 · 0 评论 -
[论文解读] Relating Code Coverage, Mutation Score and Test Suite Reducibility to Defect Density
Relating Code Coverage, Mutation Score and Test Suite Reducibility to Defect Density文章目录Relating Code Coverage, Mutation Score and Test Suite Reducibility to Defect Density简介摘要介绍目标和方法实验结果Properties of Subjects变异体生成突变体分类突变分数测试缩减缺陷密度结果分析相关工作结论简介论文标题Rela原创 2020-07-12 19:55:37 · 552 阅读 · 0 评论 -
[论文解读] Beyond Code Coverage – an Approach for Test Suite Assessment and Improvement
Beyond Code Coverage – an Approach for Test Suite Assessment and Improvement文章目录Beyond Code Coverage – an Approach for Test Suite Assessment and Improvement简介摘要介绍动机和目标TAIME 方法方法度量用例对SODA的改进相关工作前景展望简介论文标题Beyond Code Coverage – an Approach for Test Suit原创 2020-07-10 19:21:06 · 332 阅读 · 0 评论 -
[论文解读]Coverage Is Not Strongly Correlated with Test Suite Effectiveness
Coverage Is Not Strongly Correlated with Test Suite Effectiveness文章目录Coverage Is Not Strongly Correlated with Test Suite Effectiveness简介摘要介绍相关工作方法论术语目标程序生成错误程序生成测试套件测量覆盖范围衡量有效性结果规模是否与有效性相关当忽略规模时,覆盖率是否与有效性相关?当规模固定时,覆盖率是否与有效性相关?讨论对有效性的威胁结构有效性内部有效性外部有效性简介论原创 2020-07-09 14:42:44 · 453 阅读 · 0 评论 -
[论文解读] Adversarial Attacks on Deep Learning Models in Natural Language Processing: A Survey
Adversarial Attacks on Deep Learning Models in Natural Language Processing: A Survey文章目录Adversarial Attacks on Deep Learning Models in Natural Language Processing: A Survey简介摘要介绍自然语言处理中的对抗性攻击与深度学习技术综述深度学习模型的对抗性攻击:一般分类法定义威胁模型。测量。深度学习在自然语言处理中的应用从图像到文本精心制作对抗原创 2020-06-18 10:50:46 · 2676 阅读 · 0 评论 -
[论文解读] DeepRoad: GAN-Based Metamorphic Testing and Input Validation Framework for Autonomous Driving
DeepRoad: GAN-Based Metamorphic Testing and Input Validation Framework for Autonomous Driving Systems文章目录DeepRoad: GAN-Based Metamorphic Testing and Input Validation Framework for Autonomous Driving Systems简介摘要介绍背景DNN体系结构方法基于DNN的自动驾驶系统蜕变测试基于DNN的自动驾驶系统的输入验原创 2020-05-31 23:09:41 · 1886 阅读 · 0 评论 -
[论文解读] Concolic Testing for Deep Neural Networks
Concolic Testing for Deep Neural Networks文章目录Concolic Testing for Deep Neural Networks简介摘要介绍相关工作DNNs的鲁棒性Concolic测试与相关工作对比深度神经网络DNNS的覆盖测试激活模式形式化测试覆盖标准测试覆盖率指标具体覆盖要求Lipschitz连续性神经元覆盖率修改条件/决策(MC/DC)覆盖范围神经元边界覆盖我们的方法概述排名覆盖要求启发法新的具体投入的符号化生成利用线性规划进行符号分析基于全局优化的符号分原创 2020-05-29 17:21:40 · 1532 阅读 · 0 评论 -
[论文解读] DeepStellar: Model-Based Quantitative Analysis of Stateful Deep Learning Systems
DeepStellar: Model-Based Quantitative Analysis of Stateful Deep Learning Systems文章目录DeepStellar: Model-Based Quantitative Analysis of Stateful Deep Learning Systems简介摘要介绍综述RNN的状态转移建模RNN内部状态和状态转换抽象状态转换模型轨迹相似性度量将训练好的RNN表示为离散时间马尔可夫链RNN的覆盖标准State-Level Covera原创 2020-05-15 15:19:14 · 1284 阅读 · 0 评论 -
[论文解读] DLFuzz: Differential Fuzzing Testing of Deep Learning Systems
DLFuzz: Differential Fuzzing Testing of Deep Learning Systems文章目录DLFuzz: Differential Fuzzing Testing of Deep Learning Systems简介摘要动机DLFUZZ方法架构算法结论结论简介论文标题DLFuzz: Differential Fuzzing Testing of Deep Learning SystemsDLFuzz:深度学习系统的差分模糊测试2018.12DLFuz原创 2020-05-14 17:23:36 · 3645 阅读 · 0 评论 -
[论文解读]Test Selection for Deep Learning Systems
Test Selection for Deep Learning Systems文章目录Test Selection for Deep Learning Systems简介摘要初步动机选择指标测试令人惊讶的充分性指标 SA模型不确定性度量实验设置目标和方法简介论文标题Test Selection for Deep Learning Systems深度学习系统的测试选择2019.4...原创 2020-04-09 15:19:12 · 761 阅读 · 1 评论 -
[论文解读] DeepCruiser: Automated Guided Testing for Stateful Deep Learning Systems
DeepCruiser: Automated Guided Testing for Stateful Deep Learning Systems文章目录DeepCruiser: Automated Guided Testing for Stateful Deep Learning Systems简介摘要背景RNNAutomated Speech Recognition覆盖率制导的测试框架Deep...原创 2020-05-08 13:56:30 · 1257 阅读 · 0 评论 -
[论文解读]NBDT: Neural-Backed Decision Trees
NBDT: Neural-Backed Decision Trees文章目录NBDT: Neural-Backed Decision Trees简介摘要初步相关工作方法使用嵌入式决策规则进行推理构建诱导层次结构用树木监督损失进行训练实验结果分析节点语义的可解释性避免准确性与可解释性的权衡树遍历的可视化实现简介论文标题NBDT: Neural-Backed Decision TreesN...原创 2020-05-02 17:17:11 · 4168 阅读 · 3 评论 -
[论文解读]DeepHunter: Hunting Deep Neural Network Defects via Coverage-Guided Fuzzing
DeepHunter: Hunting Deep Neural Network Defects via Coverage-Guided Fuzzing文章目录DeepHunter: Hunting Deep Neural Network Defects via Coverage-Guided Fuzzing摘要初步DNN软件开发和部署基于覆盖的灰盒模糊方法DeepHunter概述Transfor...原创 2020-04-29 16:58:37 · 2371 阅读 · 3 评论 -
[论文解读]TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing
TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing文章目录TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing简介摘要Background模糊过程组件实验结果CGF可以有效地发现训练好的神经网络中的数值错误CGF在模型和它们的量化版本之...原创 2020-04-27 14:17:31 · 1580 阅读 · 0 评论 -
[论文解读] Bridging Machine Learning and Logical Reasoning by Abductive Learning
Bridging Machine Learning and Logical Reasoning by Abductive Learning文章目录Bridging Machine Learning and Logical Reasoning by Abductive Learning简介摘要介绍术语Abductive LearningProblem SettingFrameworkOptimis...原创 2020-04-17 16:39:31 · 2381 阅读 · 16 评论 -
[论文解读]Structural Coverage Criteria for Neural Networks Could Be Misleading
Structural Coverage Criteria for Neural Networks Could Be Misleading简介论文标题Structural Coverage Criteria for Neural Networks Could Be Misleading神经网络的结构覆盖标准可能具有误导性2019核心adversary-oriented searc...原创 2020-04-16 15:36:00 · 603 阅读 · 0 评论 -
[论文解读] A Noise-Sensitivity-Analysis-Based Test Prioritization Technique for Deep Neural Networks
A Noise-Sensitivity-Analysis-Based Test Prioritization Technique for Deep Neural Networks文章目录A Noise-Sensitivity-Analysis-Based Test Prioritization Technique for Deep Neural Networks简介摘要内容概述噪声敏感度实例技术...原创 2020-04-14 13:53:52 · 526 阅读 · 0 评论 -
[论文解读] DeepCT:Tomographic Combinatorial Testing for Deep Learning Systems
DeepCT: Tomographic Combinatorial Testing for Deep Learning Systems简介文章目录DeepCT: Tomographic Combinatorial Testing for Deep Learning Systems简介摘要层析组合的背景和动机组合测试标准深度学习的CT稳健性测试测试生成算法实验实验结果论文标题DeepCT:...原创 2020-04-11 16:14:05 · 2037 阅读 · 0 评论 -
[论文解读] DeepMutation: Mutation Testing of Deep Learning Systems
DeepMutation: Mutation Testing of Deep Learning Systems简介论文标题DeepMutation: Mutation Testing of Deep Learning SystemsDeepMutation:深度学习系统的变异测试2018.8贡献提出了一种专门针对DL系统的变异测试框架和工作流程,实现了对测试数据集的质量评估和薄...原创 2020-04-06 16:43:41 · 2537 阅读 · 0 评论 -
[论文解读]Explaining Knowledge Distillation by Quantifying the Knowledge
Explaining Knowledge Distillation by Quantifying the Knowledge简介论文标题Explaining Knowledge Distillation by Quantifying the Knowledge可解释性:通过量化知识来解释知识蒸馏2020.3.7核心内容本研究核心在于通过定义并量化神经网络中层特征的“知识量”,从...原创 2020-03-26 21:52:26 · 1377 阅读 · 1 评论 -
[论文解读]关于机器学习测试,看这一篇论文就够了 Machine Learning Testing: Survey ,Landscapes and Horizons
Machine Learning Testing: Survey ,Landscapes and Horizons文章目录Machine Learning Testing: Survey ,Landscapes and Horizons介绍摘要简介机器学习概论机器学习测试定义ML测试工作流测试在ML开发中的作用离线测试在线测试ML测试组件ML测试属性正确性模型相关性鲁棒性安全性数据隐私效率公...原创 2020-03-22 16:45:02 · 7408 阅读 · 0 评论 -
[论文解读]Comparing the Effectiveness of Testing Techniques
Comparing the Effectiveness of Testing Techniques简介论文标题Comparing the Effectiveness of Testing Techniques比较测试技术的有效性2011.8核心内容讨论一个测试数据选择标准是否另一个更有效ppt地址本篇论文主要是讨论了几种形式化定义的测试数据集的好坏标准,以及形式化定义的缺...原创 2020-03-11 17:56:10 · 324 阅读 · 0 评论 -
[论文解读]Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey文章目录Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey简介摘要常用术语对抗性攻击用于分类的攻击Box-constrained L-BFGSF...原创 2020-03-06 22:05:11 · 4800 阅读 · 1 评论 -
[论文解读] Cross-Entropy Based Testing
Cross-Entropy Based Testing文章目录Cross-Entropy Based Testing简介摘要介绍CE 测试方法流程用途举例背景优化问题中的交叉熵方法交叉熵公式基于CE方法对图形优化问题的应用性能函数概率更新方法预备定义将交叉熵用于测试算法算法流程各部分组成改进性能的启发式应用简介论文标题Cross-Entropy Based Testing基于交叉熵的测...原创 2020-03-01 00:02:59 · 968 阅读 · 0 评论 -
[论文解读]Boosting Operational DNN Testing Efficiency through Conditioning
Boosting Operational DNN Testing Efficiency through Conditioning文章目录Boosting Operational DNN Testing Efficiency through Conditioning简介摘要DNNS的操作测试DNN测试定义DNN测试操作过程覆盖率及其困难之处高效的DNN测试方法数学定义三种采样方法SRS 简单随机抽...原创 2020-02-27 17:18:24 · 906 阅读 · 0 评论 -
[论文解读]An Efficient Evolutionary Algorithm for Subset Selection with General Cost Constraints
An Efficient Evolutionary Algorithm for Subset Selection with General Cost Constraints文章目录An Efficient Evolutionary Algorithm for Subset Selection with General Cost Constraints简介背景知识子模子集选择问题约束介绍问题目标解...原创 2020-02-20 17:12:44 · 685 阅读 · 1 评论 -
[论文解读] Callisto: Entropy based test generation and data quality assessment for Machine Learning ...
Callisto: Entropy based test generation and data quality assessment for Machine Learning Systems简介:论文标题Callisto: Entropy based test generation and data quality assessment for Machine Learning Syst...原创 2020-02-03 16:08:18 · 298 阅读 · 0 评论 -
[论文解读]Feature-Guided Black-Box Safety Testing of Deep Neural Networks
Feature-Guided Black-Box Safety Testing of Deep Neural Networks深度神经网络的功能导向黑盒安全测试总结: 一种使用蒙特卡洛树搜索的方式来获取反例的方法,使用SIFT提取关键点,并采用两个玩家回合制的方式对关键点进行操作.文章目录Feature-Guided Black-Box Safety Testing of Deep Neur...原创 2019-12-10 22:31:26 · 669 阅读 · 0 评论 -
[论文解读]Evaluate the Effectiveness of Test Suite Prioritization Techniques Using APFD Metric
Evaluate the Effectiveness of Test Suite Prioritization Techniques Using APFD Metric使用APFD度量评估测试集优先级技术的有效性背景随着软件的发展,测试用例的大小往往会增加,这使得执行整个测试用例的成本太高.为了降低回归测试的成本,要对测试用例进行优先级划分。什么是APFD?Average Percent...原创 2019-11-02 17:56:03 · 1002 阅读 · 0 评论 -
[论文解读] Testing Deep Neural Networks
简介注意:本文特别多的公式,但很多公式其实都是把一些知识用公式化的语言描述出来了而已.论文标题Testing Deep Neural Networks测试深度学习系统2018.3核心内容介绍了**三种变化,**符号变化,值变化,距离变化.介绍了四种覆盖方法ss, ds, sv, dv自动测试用例生成算法背景前人的如下两种测试方法是不好的使不同的测试样例激活不同...原创 2019-09-27 21:23:27 · 1639 阅读 · 0 评论 -
[论文解读] Guiding Deep Learning System Testing using Surprise Adequacy
简介论文标题:Guiding Deep Learning System Testing using Surprise Adequacy,使用惊讶指标SA来构建深度学习测试系统2018.8.25现状目前有两个假设第一个假设:metamorphic testing 蜕变测试, dl系统的两个输入在某些人类感觉上是相似的,那么输出也应该是相似的第二个假设,即一组输入越多...原创 2019-09-22 13:36:44 · 2011 阅读 · 4 评论 -
[论文解读] DeepGauge Multi-Granularity Testing Criteria for Deep Learning Systems
简介论文标题:DeepGauge Multi-Granularity Testing Criteria for Deep Learning SystemsDeepGauge:深度学习系统的多粒度测试标准核心内容提出了一系列的DNN的测试标准涉及到了4种对抗样本生成技术现状,背景,预备知识仅从dl输出来测量软件质量是肤浅的需要一组测试标准,而不是基于dl决策输出的...原创 2019-09-17 20:56:33 · 2777 阅读 · 1 评论