The problem is to fit a conic section to a set of *n* points
( ). A conic can be
described by the following equation:

where *A* and *C* are not simultaneously zero. In practice, we
encounter ellipses, where we must impose the constraint .
However, this constraint is usually ignored during the fitting because

- the constraint is usually satisfied if data are not all situated in a flat section.
- the computation will be very expensive if this constraint is considered.

As the data are noisy, it is unlikely to find a set of parameters
*(A, B, C, D, E, F)* (except for the trivial solution *A=B=C=D=E=F=0*)
such that , . Instead, we will try to
estimate them by minimizing some objective function.

Thu Feb 8 11:42:20 MET 1996