I am not sure how to approximately compute the mutual information in the controlled experiments on synthetic datasets.

pliang279 / FactorCL

[NeurIPS 2023] Factorized Contrastive Learning: Going Beyond Multi-view Redundancy

MIT License

55 stars 5 forks source link

I am not sure how to approximately compute the mutual information in the controlled experiments on synthetic datasets. #2

Closed sdlas closed 9 months ago

zihaod commented 10 months ago

Hi!

Thanks for your feedback. We added a section called MI Estimation in the synthetic_example notebook, which should demonstrate how to estimate MI using the Probablistic Classifier Estimator.