MATH46132/66132: Numerical Optimisation
The lecturer is Tony Shardlow. My contact
details are here.
The course
Module description
is here
and the first handout is
here.
The exams
Midsemester coursework is worth 25% and will involve MATLAB programming.
Handouts
-
First handout
-
Problem with steepest descent
-
Notes on conjugate gradient
-
Pros and cons
-
Coursework 1: worth 20%: due Friday 30th March
-
Coursework 2: worth 5%: due Friday 4th May
Problem Sheets
MATLAB code
Code for steepest descent/backtracking linesearch
algorithm: it runs with the command
inexact_sd([1.2,1.2],1e-6,1,0.001,0.8,0.9)
-
bls.m Backtracking linsearches
- f_eval.m
and g_eval.m
evaluates object fn and gradient. Note that this is not the f
given on the example sheet, because the one on the example sheet is
"too hard" for steepest descent.
-
inexact_sd.m
-
cg.m conjugate gradient for sheet 2.
-
zoom.m and
linesearch2.m for solving a lineasearch method subject to the strong
Wolfe conditions. Both codes are required and you use linesearch2.m
(zoom.m is only a subfunction).
-
GN_eg_data.m for example data.