# Algorithms for Matrix Estimation

Many applications consists of data representable in the form of a sparse and dense matrix. The task of matrix estimation involves prediction from partially or fully observed (potentially noisy) matrices entries. The structural assumptions on the underlying matrix often lead to elegant algorithms and interpretations. This project involves novel algorithms for matrix estimation problems with strong statistical guarantees. The application focus of this project is to develop a framework for image quality estimation over a network, wherein interactions between the nodes are represented in the form of a partially observed matrix. WNCG Prof. Joydeep Ghosh leads the project, with collaboration from WNCG students Suriya Gunasekar and Ayan Acharya.

This research supported by the National Science Foundation.

**Paper 1: **Exponential Family Matrix Completion Under Structural Constraints

**Paper 2: **Noisy Matrix Completion Using Alternating Minimization