I’m selected in Northeastern University-North Carolina State University Mathematics Undergraduate International Exchange Program udner the advise of Kazufumi Ito

Key Findings

  • Solving sparse optimization problems.

    When the matrix in a linear constrained problem is a sparse matrix, SVD decomposition isused. By extracting the submatrix that is greater than the tolerance value, the speed of solving using gradient descent is improved, and the numerical error of the solution is reduced.

  • SPG for solving Non-Smooth method.

    Establish a bridge between the projection gradient method and the gradient descent method. We extend the proof of convergence of the gradient descent method to the projection gradient method, and establish the RL condition for generalized gradients under strong convexity and smoothness assumptions.

The project poster

Open in the new tab for high quality image version or click here for pdf version.