It is certainly worth pursuing when you can make this
computation.
For some important classes of symbolic computation
problems, predicting the cost of doing the calculation requires
knowing more than you know at the beginning. For example,
factoring a polynomial of degree 100 can take a short time
or a long time. You will know only close to the end.
For some operations, such as computing the determinant of a
dense matrix of floating point numbers, you could predict
the time.
An alternative to this, again worth pursuing, is to have a
somewhat introspective system that says, every so often:
I'm been chewing away on this problem for 60 seconds. Maybe
I should tell the human who sent me on this errand where I
am, and ask him/her if I should continue. (Give some info
about memory usage, etc.)
In the case of factoring, one could monitor progress by
saying things like "successfully factored in a finite field."
Rarely are we able to say "we are 90% done" with a calculation,
but even so, I think the principle is important. The human
often has no idea of the expense of an operation, and could
be told as much as we can tell.
RJF
C Y wrote:
>I was wondering about something - in some cases, such as large
>operations on large matricies - it is possible to know in advance about
>how long the process will take, since one can estimate the time per
>unit operation and the total number of operations? Would it be
>possible/useful for the system to have a "sanity check" mode where it
>would try to make an estimate of the total time to complete a command,
>and if it's something long (minutes, half hour, etc.) it can ask for
>confirmation? I don't know how effective it could be, but it would
>sure be nice to know when to expect a long evaluation and when to
>suspect a mistake. I imagine this probaby is impractical, but I
>thought I'd ask :-)
>
>CY
>
>__________________________________
>Do you Yahoo!?
>Exclusive Video Premiere - Britney Spears
>http://launch.yahoo.com/promos/britneyspears/
>
>_______________________________________________
>Maxima mailing list
>Maxima@www.math.utexas.edu
>http://www.math.utexas.edu/mailman/listinfo/maxima
>
>