** Next:** Convergence Criteria
** Up:** Velocity Inversion
** Previous:** Parameter Encoding

For the fitness function I tested both an *L*_{1} and an *L*_{2} norm
of the sample-wise difference between the two traces as the criterion to measure
the fit. That is, I used
| |
(3) |

and
| |
(4) |

where *x* and *y* are the sample amplitudes of the reference and the trial trace
respectively, *j* represents the *j*th individual and *N* is the number of time
samples.
These fitness functions consider each sample in the sub-sampled well log as being
completely independent of any other. Notice that I am using ``fitness'' here
to actually mean the opposite of its usual meaning. That is, an individual will
be considered to be highly fit if its fitness value, as defined above, is very
low, i.e., it is a good match to the original trace.

** Next:** Convergence Criteria
** Up:** Velocity Inversion
** Previous:** Parameter Encoding
Stanford Exploration Project

11/11/2002