[Insight-developers] Progress behaviour

[Kurt Augustine]

Bill,
For the case of iterative filters, here's a thought, how about 1 - 1/n,
where n is the number of iterations completed.  Of coarse this will quickly
approach the "clamping model" when the number of iterations gets large so
you could condition it with your "best guess" value by raising the value to
the power of the log of your guess.  i.e. (1-1/n)^log(m), were n is the
number of iterations completed and m is the "best guess" max number of
iterations.  Actually, I think the natural log (ln) would be better here
than log: (1-1/n)^ln(m).
I believe this model will keep the progress below 1, avoids quick jumps to
the end, and reduces clamping.  If you provide a means for canceling at the
end of every iteration, this may be all you need for proper progress
functioning.  I can't say I gave this model a thorough investigation (not
really sure what our n values and m values are in reality) so I could be way
off but the numbers could be tweaked to give a better model.

If this doesn't look viable, I would prefer wrapping over clamping.

Just a thought ... Kurt


> -----Original Message-----
> From: Lorensen, William E (Research) [mailto:lorensen at crd . ge . com]
> Sent: Thursday, June 19, 2003 1:39 PM
> To: insight-dev-list
> Subject: [Insight-developers] Progress behaviour
> 
> 
> Folks,
> We are assessing the state of progress reporting in the 
> system. For the most
> part, the ProgressReporter and ProgressAccumulator work great.
> 
> The ProgressReporter requires a guess for the number of pixels to be
> visitied.
> 
> The trouble appears in filters that are iterative. These 
> filters seldom know
> how many iterations will be performed. Sometimes we can make 
> a guess based
> upon worse cases or averages.
> 
> Two results are possible when the guess is wrong. If we guess too many
> iterations, the progress may jump from a low number to 1 at 
> the end of the
> filter. If we guess too few, the progress can exceed one. The 
> later is not
> acceptable. We can fix this by clamping or wrapping.
> 
> What do people prefer:
> 
> 1) Jumping: progress increments, jumps to end
> 2) Wrapping: progress increments, if it reaches 1, drops back.
> 
> Bill
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at www . itk . org
> http://www . itk . org/mailman/listinfo/insight-developers
>