Numerical Linear Algebra

Postgraduate course

Course description

Objectives and Content

This course deals with algorithms to solve: The eigenvalue problem superdeterminant equation systems and linear equation systems (onlyKrylov subspace iteration). Algorithms for matrix decomposition as QR-factorisation and Singular-value decomposition will be discussed and analysed with respect to stability and complexity.

Learning Outcomes

Upon completion of the course, the successful candidate shall be able to:

  • Choose the most appropriate numerical method to solve a given linear algebra problem.
  • Explain and describe the principles of SVD, QR, LU and Cholesky factorization of matrices.
  • Have the knowledge of different methods for eigenvalue problems, like power method, divide and conquer and QR method.
  • Explain the principles of Krylov subspace methods, like the Arnoldi iteration, GMRES, Lanczos iteration and conjugate gradients.
  • Analyze the speed and rate of convergence and stability of numerical algorithms.

Semester of Instruction

Autumn
Required Previous Knowledge
None
Recommended Previous Knowledge
MAT160 Scientific computing I
Credit Reduction due to Course Overlap
INF261: 10 ECTS
Compulsory Assignments and Attendance
Exercises.
Forms of Assessment
Oral exam. Exercises might be graded and included in the final grade.
Grading Scale
The grading scale used is A to F. Grade A is the highest passing grade in the grading scale, grade F is a fail.