Non-linear equations and numerical optimization

AA 2004/2005

Lessons of 21/3/2005 and 22/3/2005

One Dimensional Non-Linear Problems

screen version slides-1D-1x1.pdf

print version 2 slides by sheet slides-1D-1x2.pdf

print version 4 slides by sheet slides-1D-2x2.pdf

Lessons of 4/4/2005 and 5/4/2005

One-Dimensional Minimization

screen version slides-m1D-1x1.pdf

print version 2 slides by sheet slides-m1D-1x2.pdf

print version 4 slides by sheet slides-m1D-2x2.pdf

Unconstrained minimization

screen version slides-mND-1x1.pdf

print version 2 slides by sheet slides-mND-1x2.pdf

print version 4 slides by sheet slides-mND-2x2.pdf

Lessons of 18/4/2005 and 19/4/2005

Conjugate Direction minimization

screen version slides-CG-1x1.pdf

print version 2 slides by sheet slides-CG-1x2.pdf

print version 4 slides by sheet slides-CG-2x2.pdf

Lessons of 2/5/2005 and 3/5/2005

Non-linear problems in N variable (preliminary draft)

screen version slides-ND-1x1.pdf

print version 2 slides by sheet slides-ND-1x2.pdf

print version 4 slides by sheet slides-ND-2x2.pdf

Quasi-Newton methods for minimization (preliminary draft)

screen version slides-mQN-1x1.pdf

print version 2 slides by sheet slides-mQN-1x2.pdf

print version 4 slides by sheet slides-mQN-2x2.pdf

Lessons of 9/5/2005 and 10/5/2005

Trust Region Method (preliminary draft)

screen version slides-TR-1x1.pdf

print version 2 slides by sheet slides-TR-1x2.pdf

print version 4 slides by sheet slides-TR-2x2.pdf

Final exam task list

  • The Preconditioned Conjugate Gradient method

  • The Fletcher Reeves and Polack Ribiere method

  • The Broyden method for system of nonlinear equations

  • The SR1 method for uncostrained optimization

  • The DFP and BFGS method for uncostrained optimization

  • The double dogleg method

  • The exact trust region method

  • The Levemberg-Marquardt method

  • The Sequential Quadratic Programming (SQP) method

  • The multidimensional Newton method and the Kantorovich proofs of its local quadratic convergence

The candidate must choose an argument and perform the following task:

  • Search in the literature to obtain a list of significative reference (10-20 reference)

  • Write a report describing the argument

  • Write a C++ code implementing the algorithm

The code should be commented in the report, or in alternative well documented using e.g. DOXIGEN.