search

YyzHarry / imbalanced-regression

[ICML 2021, Long Talk] Delving into Deep Imbalanced Regression
http://dir.csail.mit.edu
MIT License
819 stars 129 forks source link

Applying LDS/FDS to classic machine learning models #31

Closed luopx closed 2 years ago

luopx commented 2 years ago

Hi! This work is really fantastic! However, I found it hard to apply LDS/FDS to classic machine learning models like random forest. For example, after getting the effective label density with LDS, how should I use this?

YyzHarry commented 2 years ago

Hi, thanks for your interest. For LDS and FDS, they should be easy to apply to classic machine learning models.

For LDS, since it operates on the label space, it is independent of the model you use. For example, if using random forest (regressor), referring to the fit() method of the sklearn RandomForestRegressor (and other classic methods like SVM), there is an argument sample_weight, where you can provide the weight of each input sample. You can then use the effective label density by LDS to compute the weights.

As for FDS, it might be less straightforward; but since it only needs feature statistics, if the algorithm can generate intermediate features for each input, it can also be adapted, depending on the actual method you use.

luopx commented 2 years ago

Thank you very much for your reply! This really helped me a lot!