search

nmichlo / disent

🧶 Modular VAE disentanglement framework for python built with PyTorch Lightning ▸ Including metrics and datasets ▸ With strongly supervised, weakly supervised and unsupervised methods ▸ Easily configured and run with Hydra config ▸ Inspired by disentanglement_lib
https://disent.michlo.dev
MIT License
124 stars 18 forks source link

[FEATURE]: Add IRS Metric #34

Open ema-marconato opened 2 years ago

ema-marconato commented 2 years ago

Hello, it would be nice to integrate also Interventional Robustness Score (IRS). https://arxiv.org/abs/1811.00007

Do you believe it is easy to integrate?

Bests

nmichlo commented 2 years ago

Hi there,

I can't be sure right now, would need to sit with the paper for a while and evaluate.

Do you know of any existing implementations that can be used as a reference? Also, is there any chance you would you be willing to contribute?

ema-marconato commented 2 years ago

Hi, yes you can find an implementation here: https://github.com/google-research/disentanglement_lib. I don't have time now to work on it, but there 3/4 interesting causal measures of disentanglement that are very useful. Let you know if I have time for that. Thank you for your work. Cheers,

Emanuele