In our privacy course, it was mentioned that in the report noisy max algorithm, we define a set of outcomes. Then, we will compute their scores and add a noisy sampled for example from a Laplace distribution.
I want to ask whether the set of outcomes can be anything. For example, could I do a SQL conditioned query "SELECT COUNT(*), Status FROM table WHERE Gender = "Female" GROUP BY Status", treat the possible status as options, consider the count as score and apply report noisy max?
I researched on the internet.But I didn't find much except the Conditional differential privacy. For this reason, I assume it remains epsilon differentially private.
