CIS 610, Summer 1, 2009
Brief description:
The purpose of this course is to present geometric methods used in
computer vision, robotics and computer graphics.
The treatment will be rigorous but
I will try very hard to convey intuitions and to give many
examples illustrating all these concepts.
Syllabus:
This time, I intend to discuss two kinds of topics:
-
Affine and Euclidean geometry, spectral theorems, SVD, pseudo-inverses and
PCA,
-
Convex sets, polytopes, polyhedra and basics of linear programming
The material in (2) provides firm foundations for
linear programming and convex programming.
-
Basics of Affine Geometry
-
Affine spaces, affine combinations (barycenters)
-
Affine subspaces, affine independence, affine frames,
convex combinations
-
Affine maps, affine groups
-
A glimpse at affine geometry
(Pappus, Desargues)
-
Hyperplanes
and affine forms, intersection of affine spaces.
(Chapter 1-2 of GMA)
-
Euclidean Geometry
-
Inner Products, Euclidean Spaces
-
Orthogonality, Orthogonal and Orthonormal bases, Gram-Schmidt
-
A glimpse at Fourier Series
-
Linear Isometries (also called orthogonal transformations),
the orthogonal group, orthogonal matrices,
the Groups O(n), SO(n)
-
Hyperplane reflections and the Cartan-Dieudonne' Theorem
-
QR-decomposition
-
Householder matrices and QR-decomposition
(Chapters 6 and 7 of GMA)
-
Spectral theorems in Euclidean geometry and Hermitian spaces,
-
Normal linear maps, self-adjoint linear maps, skew self-adjoint linear maps
(Chapter 11 of GMA)
-
Singular value decomposition and polar forms
-
Polar form
-
Singular value decomposition (SVD)
-
Pseudo-inverses
-
Least squares and PCA
(Chapters 12 and 13 of GMA)
-
Convex sets and Polytopes
-
Basic properties of convex sets,
Half spaces determined
by hyperplanes
-
Caratheodory's theorem on convex hulls
-
Radon's theorem
-
Helly's theorem.
-
Existence of center points
-
Separation theorems:
Hahn-Banach theorem and others.
-
Supporting hyperplanes, Minkowski's theorem.
-
The Farkas lemma
-
Vertices and Extremal points, Krein and Milman's theorem.
-
Polytopes and polyhedra: H-polytopes and V-polytopes. Cones.
Vertices, faces and facets.
-
Polarity and duality.
The main theorem.
-
Fourier-Motzkin elimination, the equivalence of H-polyhedra and V-polyhedra
revisited
We may also discuss applications to the Support Vector Machine Method
(Chapter 3 of GMA + Notes)
-
Foundations of Linear Programming
-
Linear programming
-
Primal and Dual Programs
-
Lagrangian functions and
Strong Duality
-
Application to fractional packing and covering problems
-
Introduction to primal-dual combinatorial algorithms
Back to
Gallier Homepage
published by: