Memory-efficient structured convex optimization

This thesis focuses on the development of memory-efficient algorithms for a class of mathematical optimization problems. The problem consists of finding the best possible value of the decision variable that optimizes an objective function over a given set of all possible solutions. We exploit the key properties of the problem to develop a new technique to represent the decision variable. With this method, we can potentially reduce the storage requirements for solving large-scale instances of problems encountered in a wide range of application areas such as control theory, combinatorial optimization, and machine learning.