[
https://issues.apache.org/jira/browse/MATH-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13920923#comment-13920923
]
Gilles commented on MATH-1105:
------------------------------
bq. Since the number of accepted observations can change from one Evaluation to
the next, I couldn't just use LSP.getObservationSize(). Since the current
optimizers aren't set up to support a changing observation size, I didn't want
to change the dimension of the residuals or the Jacobian either.
That's a key design issue.
Does it makes sense for the "LeastSquaresProblem" instance to change the number
of observations? According to the feature discussed here, it seems that the
answer could be yes. If so, the optimizer cannot assume that the number of
observations is constant.
On the other hand, if we consider that changing an optimizer code to allow for
a changing number of observations is not standard practice, that should be
explicitly reflected in the handling of bad data, i.e. no observation can be
removed, only its weight can be altered.
Then, "getAcceptedObservations()" is not enough to convey the transformation
which the handler applied on bad data; only the weights differences (between
the original and the wrapper instances) will provide the full story. Hence, it
seems that adding this method complexifies the API without a clear benefit.
> Least squares statistical data editing
> --------------------------------------
>
> Key: MATH-1105
> URL: https://issues.apache.org/jira/browse/MATH-1105
> Project: Commons Math
> Issue Type: Improvement
> Reporter: Evan Ward
> Attachments: 0001-Add-statistical-editing-capability.patch,
> 0002-Integrate-data-editing-with-the-LS-framework.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
