Is it ok to try to revive this posting?

It would be a great help to me if fast_float() could be gotten to work
on large polynomials (>60,000 terms).

I suppose that fast_float() makes a space-time tradeoff  that might be
insurmountable.  Still, the fact that fast_float(p) kills the
worksheet namespace when p is a polynomial with >60,000 terms suggests
the possibility of a bug.

So far, sage/numpy has succeeded in minimizing a 4th degree polynomial
in 100 variables containing 1,000,000 terms in 18,000s cpu time, after
using fast_float() on each component of the gradient before supplying
the gradient to minimize().

For larger problems, fast_float() fails on the components of the
gradient.  Without using fast_float(), sage/numpy succeeded in
minimizing a 4th degree polynomial in 120 variables containing
2,000,000 terms in 145,000s cpu time.

I'm presently trying to minimize a 4th degree polynomial in 144
variables containing 4,000,000 terms, again without fast_float().   I
can't be sure that fast_float() would give this calculation any better
chance of finishing within my lifetime, but it would at least lift my
hopes.

thanks,
Daniel Friedan

On Jul 20, 11:00 am, Daniel Friedan <[email protected]> wrote:
> fast_float(p) trashes the worksheet namespace when p is a polynomial
> with 60,000 terms.  I tried to pin down the minimum number of terms
> that produce this bug, arriving at 56093, but the bug was not
> consistent.   (Sage 4.3.4 under Linux, Sage 4.4.4 under OS X 10.6.4)
>
> sage: Nterms = 60000
> sage: RQ = PolynomialRing(QQ, 1, 'x', sparse=True)
>          RQ.inject_variables(); vars = RQ.gens(); var=vars[0]
>
>          Defining x
>
> sage: p=0;
>          for k in range(Nterms):
>               p += var^k;
>          p in RQ
>
>         True
>
> sage: len(p.coefficients())
>
>         60000
>
> sage: g=fast_float(p)
> sage: Nterms
>
>           Traceback (click to the left of this block for traceback)
>            ...
>            NameError: name 'Nterms' is not defined

-- 
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/sage-support
URL: http://www.sagemath.org

Reply via email to