ryanwangzf/influence_subsampling

Official Implementation of Unweighted Data Subsampling via Influence Function - AAAI 2020

Hi there 👋 This is Zifeng.

I am a PhD student of computer science at UIUC. Please refer to my homepage for more about me.

influence_subsampling's Issues

What are the simplest methods for the label noise problem?

If I have enough low quality data from unsupervised methods or rule-based methods.

I read from https://github.com/subeeshvasu/Awesome-Learning-with-Label-Noise ,but these methods are a little complex for me.

In detail, I deal with a multi-label classification task. First I crawl web page such as wiki and use regex-based rule to mark the label. The model input is the wiki title and the model output is the rule-matched labels from wiki content. My task is to predict the labels for the wiki title.

Do you think removing the wrong data predicted by trained model is a simple but effective method?

@RyanWangZf Thank you very much!

The dataset about Cifar10 or SVHN

Dear Zifeng Wang
My name is Kong shuming and I'm a graduate student in ShangHai JiaoTong universitry. Recently I'm interested in your paper "Less Is Better: Unweighted Data Subsampling via Influence Function". I totally agree with you that sometimes using the smaller data via influence function will get better performance. I get the same result in MNIST dataset and cancer dataset. I follow your github and try to process the dataset such as cifar10 or SVHN. However I am confused about how to create the raw text of the dataset. So I am really appreciated if you could show me one example or give me the dataset.npy. So one line in the raw text is label plus feature vector? Can you give me one example such as cifar10 raw text? Thank you all the time. My email is [email protected].

                                                                                                                                                                                                                                                                           Best wishes                                                                                                                                                                                                                                                                         Kong shuming                                                                                                                                                                                                                                                                               2020/3/29                                                                

tabular data/ noisy instances/ new datasets

Hi,
thanks for sharing your implementation. I have some questions about it:

1. Does it also work on tabular data?
2. Is the code tailored to the datasets used in the paper or can one apply it to any data?
3. Is it possible to identify the noisy instances (return the noisy IDs or the clean set)?

Thanks!