Can’t we use Bayes Rule to determine P(X|Y) of the generator from a trained classifier ?

Hi @chams

You can! only if you have the necessary probabilities.

If you have a trained classifier that predicts class labels given data generated by a generator, you can estimate P(Y∣X). However, to use Bayes rule to estimate P(X∣Y), you would also need to know P(X) (if you have prior information or a large dataset to estimate its distribution).

Hi @Alireza_Saei,

Thank you for your response. Can i also also ask how effective is it compared to GANs ? And is it commonly used ?

Bayes Rule for estimating P(X∣Y) from a classifier can be effective with accurate prior information, but it’s limited by the need for data. Compared to GANs (which excel in generating realistic data without explicit priors), Bayes Rule is less commonly used due to its reliance on prior knowledge and its limitations in capturing complex data distributions.