When the stack construction phase is complete, QEPCAD has a CAD of
free-variable space with truth values attached to each cell. This
represents the solution set of the (possibly quantified) input
formula. The last step is to represent this set as a formula (a
solution formula), and
one of the big advantages of quantifier elimination by CAD is its
ability to produce simple solution formula. If you
simply type go at the "Before Solution >"
prompt, QEPCAD will automatically construct a (hopefully) simple
solution formula in the language of Tarski formulas. However,
more options are available with the solution-extension
command. To understand these options a bit better it's good to
know a little bit about how solution formulas are constructed.
QEPCAD constructs solution formulas from the polynomials in the
projection factor set defining the CAD of free-variable space.
It is not always possible to construct a Tarski formula solely
from these polynomials, in which case we must add polynomials to
the projection factor set.
A CAD-with-truth-values for which a Tarski solution formula can
be built exclusively from the elements of the projection factor
set is called projection definable.
The command "pdq", which stands for projection
definability query, allows the user to ask QEPCAD whether the
CAD in the solution formula construction phase is projection
definable. It is, however, always possible
to construct an Extended Tarski
Formula
(ETF) solution
formula for a CAD-with-truth-values using just the polynomials
in the projection factor set.
The Basic Options
The solution-extension command takes a letter as an
arguments that provides several different possibilities for
solution formula construction.
Enter an informal description between '[' and ']':
[Produces a CAD that's not projection definable ]
Enter a variable list:
(x,y)
Enter the number of free variables:
1
Enter a prenex formula:
(Ey)[ x^2 + y^2 - 3 = 0 /\ x + y > 0 ].
=======================================================
Before Normalization >
go
Before Projection (y) >
go
Before Choice >
go
Before Solution >
pdq
CAD is not projection definable.
Before Solution >
solution-extension T
An equivalent quantifier-free formula:
x^2 - 3 <= 0 /\ [ x >= 0 \/ 2 x^2 - 3 < 0 ]
Before Solution >
solution-extension E
An equivalent quantifier-free formula:
x^2 - 3 <= 0 /\ x > _root_1 2 x^2 - 3
solution-extension T
This option produces a formula in the language of Tarski
formulas, and in fact this exactly the routine that gets
called if you simply type "go" at the
"Before Solution >" prompt, or
"finish" at any point. If the CAD is not
projection definable, polynomials will be added to the
projection factor set.
solution-extension E
This option produces a formula in the language of
Extended Tarski Formulas
(ETFs). Projection factors are never added to the projection
factor set.
solution-extension G
This option tries to produce a formula that gives insight into
the geometry of the solution set. The method of doing so, and
the properties of the formulas that the method produces are
described in Chapter 8 of
my thesis.
Generally the formula will be an ETF rather than a Tarski
formula. The basic idea is that the set gets described in
diskoint pieces.
solution-extension F
This option produces a formula based purely ono the
full-dimensional cells in free variable space. This means
that the truth values of lower dimensional cells are ignored.
solution-extension I
This option enters the user into an "interactive" mode, which
gives him more control over how solution formulas are
constructed.
G" option,
because it really isn't interesting for CAD's of 1-space - it
needs at least 2-levels to do anything fun. You see in this
example that the polynomial x only appears in the formula
produced by "soltution-extension T". That polynomial
is not in the projection factor set, and was actually added after
the fact by QEPCAD in order to make it possible to construct a
solution formula in the language of Tarski formulas.