New edition of my probability book with Jessica Hwang is out! Read it FREE online at probabilitybook.net

How should we implement the exponential mechanism for differentially private computation of a continuous function? All the resources I have seen yet are focused on discrete output space. I would appreciate any hint on this question. Differential Privacy OpenDP

It's unsettling how the hot topics just come and go like nothing in ML (or is it AI now): manifold learning, kernels, GANs, RL, conformal calibration... You know what doesn't? Foundations. In the best papers, the method is a vehicle for bigger ideas that stay timeless.