Seminar: Differential Privacy, Inference, and Optimisation by Sinan Yıldırım


 Differential Privacy, Inference, and Optimisation

Sinan Yıldırım

Faculty of Engineering and Natural Sciences, Sabancı University



Data privacy is an ever-growing issue in the modern world. My talk will begin with an introduction of a notion of data privacy, namely differential privacy, which is followed by two parts. In the first part, I will present a novel differentially private Markov chain Monte Carlo (MCMC) algorithm whose target distribution is the posterior distribution conditioned on the private data. We also show that, in a model with independent observations, this MCMC algorithm has desirable convergence and privacy properties that scale with data size.In the second part of the talk, I will mention some of our findings that address the question of tuning a differentially private gradient-based algorithm for optimisation.

The first part of the talk was conducted in collaboration with Beyza Ermiş. The second part is an ongoing work, conducted in collaboration with Nurdan Kuru, İlker Birbil, and Mert Gündüzbalaban.


Short Bio:

Dr Sinan Yıldırım is a faculty member in Engineering and Natural Sciences at Sabancı University, Turkey, since September 2015. He received his BS and MS degrees in Electrical and Electronic Engineering at Boğaziçi University, Turkey, between 2002 and 2009. He holds a PhD degree in Mathematical Statistics from the University of Cambridge, UK, where he studied between 2009 and 2013. He then worked under the EPSRC project "Intractable Likelihood: New Challenges from Modern Applications" as a postdoctoral researcher at the University of Bristol from 2013 to 2015. His primary research areas are Bayesian Statistics and Monte Carlo methods. More information can be found in


All interested are cordially invited.  


DATE   : Friday, February 21, 2020
TIME    : 15:00-16:00
ROOM  : VYKM 2, 5th floor of Engineering Building (Perkins Hall)