Dissertation Title: Connecting Silos for Distributed and Private Computation

Abstract: 

Data in today's world is increasingly siloed across a wide variety of entities with varying resource constraints. The quality of wisdom generated from a collaborative processing of such data is substantially better if the data from all these entities is shared across each other or centralized at a nodal entity. Such data sharing and centralization is often prohibited due to stringent privacy regulations, computational constraints, communication bottlenecks, trade secrets, trust issues and competition. This necessitates development of efficient methods for distributed computation while preserving privacy to generate wisdom whose quality is on par with the case of data centralization. This defense covers methods that I introduced for the same in an inter-disciplinary manner to tackle several such problems in distributed and private computation. The methods introduced in the thesis can be categorized into three parts including Part I.) Distributed and Private Statistical Inference (DPSI), Part II.) Distributed and Private Machine Learning (DPML) and Part III.) Distributed and Private Scientific Computing (DPSC) to benefit various downstream applications in this setting for a broader social impact at scale.

Committee members: Ramesh Raskar, Associate Professor, Massachusetts Institute of Technology (MIT)
Alex Pentland, Professor, Massachusetts Institute of Technology (MIT)
Kun Zhang , Associate Professor, Carnegie Mellon University (CMU) and MBZUAI
Han Yu, Assistant Professor, Nanyang Technological University (NTU)