Xuemei Liu is an undergraduate student at University of South Australia. Currently, she is majoring in Mathematical Sciences. She is particularly interested in the area of Optimization.
Study Of Duality Schemes With Sequential Lagrangian Updates
Augmented Lagrangian schemes are used for solving non-smooth and non-convex problems optimization problems. It is known that using a constant matrix in the linear term of the augmented Lagrangian expression leads to convergent primal dual schemes. This raises the question of whether an iterative scheme in which this matrix is updated at each iteration can still generate a convergent scheme. We will investigate different update rules in which the matrix used in the linear term of the Lagrangian is changed at each iteration. Our aim is to determine the correct assumptions needed for convergence. The supervisor provides guidan