(TENTATIVE) OUTLINE OF CBMS LECTURES BY BERND STURMFELS
Lecture 1: Polynomials in One Variable
Basic properties of univariate polynomials, companion
matrices,
Fundamental Theorem of Algebra, solvability in terms of radicals,
Sturm sequences, Descartes' rule of signs, numerical root finding.
Lecture 2: Hypergeometric Functions
We study the roots of the general polynomial of degree n
as an algebraic function of all n+1 coefficients, and we
determine the system of linear differential equations
which characterizes this algebraic function. This leads to
2^{n-1} distinct Puiseux series formulas for the roots.
Lecture 3: Groebner Bases and Numerical Linear Algebra
We show how Groebner bases for any term order can be
used to solve zero-dimensional polynomial systems, by
setting up an appropriate eigenvalue problem. The trace
form can be used to count the number of real roots.
Lecture 4: Rational Univariate Representation
Faugere and Roullier have the currently best Groebner based
method for finding all complex (and all real) roots of a
polynomial system. One key idea is that of a rational univariate
representation of the solution set. We explain how this works,
and we discuss the state of the art on Groebner bases methods
for solving zero-dimensional systems.
Lecture 5: Resultants
Resultants are very useful for studying square systems, where the number
of equations equals the number of variables. We give summary of known
exact determinantal formulas for resultants, specifically for dense
systems in few variables. This includes references to Chow forms and
sparse resultants.
Lecture 6: Homotopy Methods
We give a survey of numerical homotopy methods for solving polynomial
equations, including sparse and determinantal systems, including
recent work of Li, Sommese, Verschelde, and their collaborators.
Lecture 7: The Real Nullstellensatz
The real Nullstellensatz is a common generalization of the
usual Nullstellensatz (over an algebraically closed field)
and Linear Programming Duality (for linear inequalities).
It says that a polynomial system over the reals either has
a solution or there exists an obvious certificate for its
non-solvability. A special case concerns the representability
of positive polynomials as sums of squares.
Lecture 8: Polynomial Optimization and Semidefinite Programming
We discuss the problem of minimizing real polynomial functions,
possibly subject to semi-algebraic constraints. We show how
the highly efficient technique of semi-definite programming
can be used to solve such problems.
Lecture 9: Primary Decomposition
Solving a polynomial system means to decompose the solution set
into irreducible components, or more generally into its primary
components. We show how this can be done, first for the simpler
cases of monomial and binomial ideals, and then for general
polynomial ideals.
Lecture 10: The Palamodov-Ehrenpreis Theorem
A system of polynomial equations can be regarded as a linear
system of differential equations with constant coefficients.
In this lecture we explain what can be gained by this point
of view. We argue that the Palamodov-Ehrenpreis Theorem, a
classical result in analysis, can be seen as the ultimate
representation of the solutions to a polynomial system.
In particular, we discuss recent work of Oberst which
makes the Palamodov-Ehrenpreis Theorem effective.