PS: Deep Neural Networks Are Easily Fooled__

Deep Neural Networks Are Easily Fooled: High Confidence Predictions for Unrecognizable Images full paper
Anh Nguyen, Jason Yosinski, Jeff Clune

这篇paper的idea我很喜欢，不久的将来基于这篇paper还会产生一些工作。

这篇文章针对的问题是这样的。虽然现在DNN已经在主流分类数据集上已经展示了比传统方法更好的分类性能，但是论文的方法揭示了，我们可以轻而易举的产生一些欺骗DNN的图片：人类看不清是什么，但DNN以99.99%的confidence分类成某一个类别。

在DNN近年来一路唱红的时代，一些没有节操的媒体(当然，还有一些没节操的researcher)说可以去beat human vision。每当看到这个都只能说，呵呵。

好吧，扯多了，说一些细节。论文用了两种方法，一种采用遗传算法，还有一种采用梯度上升。这两种方法都产生出了一张有规律或无规律的图形，但DNN轻易分错的实例。近年来，也有一些paper来探讨类似的问题。具体细节看论文吧，discussion非常值得一看。

PS: Deep Neural Networks Are Easily Fooled___CVPR2015相关推荐

Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
论文原文:Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images ...
[论文阅读笔记]Deep Neural Networks are Easily Fooled:High Confidence Predictions for Unrecognizable Images
Deep Neural Networks are Easily Fooled:High Confidence Predictions for Unrecognizable Images(CVPR201 ...
论文-Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images Deep N ...
[深度学习论文笔记][Adversarial Examples] Deep Neural Networks are Easily Fooled: High Confidence Predictions
Nguyen, Anh, Jason Yosinski, and Jeff Clune. "Deep neural networks are easily fooled: High conf ...
对抗样本论文学习：Deep Neural Networks are Easily Fooled
近日看了一些对抗样本(adversarial examples)方面的论文,在这里对这些论文进行一下整理和总结. 以下仅代表个人理解,本人能力有限难免有错,还请大家给予纠正,不胜感激.欢迎一起讨论进步 ...
Deep Neural Networks are Easily Fooled:High Confidence Predictions for Unrecognizable Images
在卷积神经网络如日中天的现在,重要会议上的论文自然成了广大学者研究的对象.苦苦寻觅,然而并不能搜到"大家们"对论文的见解.痛定思痛,决定对自己看过的论文写点小感.只是个人看法,如有 ...
[转载]Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
论文内容总结首次提出假反例攻击,即:生成人类无法识别的图片,却能够让神经网络以高置信度分类成某个类别使用了多种不同的方法生成图片,称为"fooling images" 普通的E ...
[论文笔记]Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
作者:Anh Nguyen, Jason Yosinski, Jeff Clune 链接:https://arxiv.org/pdf/1412.1897.pdf 摘要: 本文的工作基于Christia ...
Deep Neural Networks are Easily Fooled High Confidence Predictions for Unrecognizable Images
Layer-Wise Data-Free CNN Compression 我们的无数据网络压缩方法从一个训练好的网络开始,创建一个具有相同体系结构的压缩网络.这种方法在概念上类似于知识蒸馏[23],即 ...

PS: Deep Neural Networks Are Easily Fooled___CVPR2015

PS: Deep Neural Networks Are Easily Fooled___CVPR2015相关推荐

最新文章

热门文章