• Next semester, my university is offering two math courses: one named ‘Linear & Non-Linear Programming’ and the other named ‘Applied Matrix Theory’."
  • Description of the first course:
    • “The course covers the formulation of linear programs, basic properties of linear programs, and the simplex method. It includes duality, necessary and sufficient conditions for unconstrained problems, minimization of convex functions, methods for solving unconstrained problems, equality and inequality constrained optimization, the Lagrange multipliers theorem, the Kuhn-Tucker conditions, and methods for solving constrained problems.”
  • Description of the second course:
    • “This course includes a review of the theory of linear systems, eigenvalues and eigenvectors, the Jordan canonical form, bilinear and quadratic forms, matrix analysis of differential equations, and variational principles and perturbation theory. Topics include the Courant minimax theorem, Weyl’s inequalities, Gershgorin’s theorem, perturbations of the spectrum, vector norms and related matrix norms, and the condition number of a matrix.”

​

I want to specialize in deep learning. Could anyone advise which course would be more beneficial for this field?

  • serge_cell@alien.topBEnglish
    1·
    1 year ago

    You can not do anything meaningful in ML without understanding eigenvalues, eigenvectors and norms. First course is generally useful for many practical non-DL tasks though, especially sensors fusion and computer vision (both DL and non-DL)