Student: Yue Zhang
Advisor: Weihong Guo, Associate Professor, Case Western Reserve University
Abstract: Sparse modeling aims to discover and leverage the key factors and hidden structures from the overwhelmingly large data in the modern world. The study of sparsity phenomena has become a blooming subject in mathematics, statistics,computer science and bioinformatics in the past three decades. In this presentation, we study the problem of learning data adaptive over-complete bases such that each data example can be linearly represented by few basis vectors. To tackle the computational challenges when the input data size grows dramatically, we design a novel distributed dictionary learning algorithm which generalizes the classical consensus settings. The local correctness of this algorithm is well justified under suitable assumptions. We will also review some previous results of sparse modeling in image reconstruction and feature selection as well as their guarantees.