Comments (4)
Hello authors, first of all, thank you very much for your paper and code contributions, which have been of considerable help to my research.
However, at present, our application scenario is that the training set and the test set are set to the same requirements. For example, if the training set is randomized by 5% and class-wise noise is added, then the test set must also be randomly 5% with class-wise noise.
I'd like to ask the author Orz about how to add protective noise and still have an unlearnable effect in the above scenario.
(Any ideas or suggestions will be of great help to me QAQ)
from unlearnable-examples.
Hi,
这是个好问题。我认为这个场景下还是算不可学习的,因为模型学到的是噪声而不是实际数据里的内容。对于测试来说的话,加了class-wise噪声的数据应当被认成对应的class。
from unlearnable-examples.
作者您好,近日我做了些实验,我将 在PoisonCIFAR10上训练好的类噪声,以不同的随机的方式加入到照片中(原本是根据图片的类别来添加噪声),以此来生成训练集P1 和 测试集P2 。 并且我期望模型在P1上训练,学习,在P2上测试会有较差的效果。
但实际的实验结果是,在10个epochs左右,模型在训练集和测试集上的准确率都已经超过80%,且随后持续稳步上升。
在上述的加噪声方式基础上,我尝试修改了加噪声的比例,但还是没能做到 预期的效果。(参数如下)
想请教下作者,我们如何调整,才有可能做到实现我们的期望结果呢?
(任何想法或建议都将是对我极大的帮助QAQ)
from unlearnable-examples.
作者您好,近日我做了些实验,我将 在PoisonCIFAR10上训练好的类噪声,以不同的随机的方式加入到照片中(原本是根据图片的类别来添加噪声),以此来生成训练集P1 和 测试集P2 。 并且我期望模型在P1上训练,学习,在P2上测试会有较差的效果。 但实际的实验结果是,在10个epochs左右,模型在训练集和测试集上的准确率都已经超过80%,且随后持续稳步上升。 在上述的加噪声方式基础上,我尝试修改了加噪声的比例,但还是没能做到 预期的效果。(参数如下)
想请教下作者,我们如何调整,才有可能做到实现我们的期望结果呢? (任何想法或建议都将是对我极大的帮助QAQ)
如果是class-wise噪声,随机加入到数据集会导致效果不如预期,可以尝试sample-wise的噪声。
from unlearnable-examples.
Related Issues (19)
- Why use custom models? Cannot reproduce with torchvision model HOT 3
- KeyError: 'train_subset' HOT 4
- A problem when training model on ImageNetMini HOT 1
- Mismatch of the training data augmentation between QuickStart.ipynb and main.py HOT 1
- Two problems in training code of ImageNetMini HOT 1
- A problem with bi-level optimization in the article HOT 6
- Several questions about this article HOT 4
- Some questions about training Inception-ResNet HOT 11
- Questions about training casia-webface dataset HOT 1
- 关于噪声处理的问题? HOT 4
- Some questions about face recognition poisoning attack HOT 5
- A problem about noise generating. HOT 1
- keyerror报错 HOT 1
- Generating examples using CelebA HOT 1
- Can you share your experience with fast-autoaugment HOT 2
- bugs when generating sample wise perturbation HOT 1
- About the plot setting in the paper HOT 2
- About visualizing the results according to log file HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from unlearnable-examples.