Plans for the week of March 3-7

Dear all, welcome back to FYS4411/9411.

This week we will try to wrap up our discussion of optimization methods, repeating some of the basic approaches and then discussing so-called quasi-Newton methods, which aim at circumventing the calculation of the second derivative of energy.

The plan is as follows:

Gradient methods:

1. Semi-Newton methods (Broyden's algorithm and Broyden-Farberg-Goldberg-Shanno algorithm)

2. Steepest descent and conjugate gradient descent

3. Stochastic gradient descent and variants thereof

4. Automatic differentiation

5. Discussion of implementations and work on codes

Teaching Material, videos and written material.

1. These lecture notes

2. Recommended background literature, Convex Optimization by Boyd and Vandenberghe. Their lecture slides are very useful (warning, these are some 300 pages).

The jupyter-notebook is at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/pub/week7/ipynb/week7.ipynb

Best wishes,

Morten