search

HanxunH / Active-Passive-Losses

[ICML2020] Normalized Loss Functions for Deep Learning with Noisy Labels
MIT License
129 stars 28 forks source link

tabular data/ noisy instances #4

Closed nazaretl closed 2 years ago

nazaretl commented 2 years ago

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

  1. Does it also work on tabular data?
  2. Is it possible to identify the noisy instances (return the noisy IDs or the clean set)?

Thanks!

HanxunH commented 2 years ago

Hi,

  1. It should be able to work on tabular data. There is no constraint on the types of input data it works on.
  2. We did not test such an idea. But you may start by looking into the loss values for individual samples. Higher values may reflect that they could be potentially noisy.
nazaretl commented 2 years ago

I see, thank you! Last question: your code is not tailored to the few datasets you test, right? It seems that only dataset.py contains special datasets and hard-coded parameters.

HanxunH commented 2 years ago

There are some hard-coded parameters in the dataset.py for asymmetrical noise. For noise generations on your dataset, you may need to change them.

nazaretl commented 2 years ago

thank you!