search

lalitkumarj / OMSCategory

Sage Category implementation of OMS
1 stars 1 forks source link

Problems in hecke matrix #61

Open rharron opened 11 years ago

rharron commented 11 years ago

I just uploaded a couple of scripts that crash on the hecke matrix part: p_3_N_19.sage and Hecke_polynomials_p_3_prime_level_up_to_17.sage. For the latter, the code works fine until level 17. In both instances, it is in trying to find a linear relation that it fails.

rharron commented 11 years ago

So, for the case of p=3, level 17 in the second example, here's a more precise symptom of the problem: (in the case I ran it) when Up is applied to the second basis vector, it can't find a linear relation among the 4 vectors (the three basis vectors and Up times the second basis vector), up to the desired precision. If I hack it to only look for a linear relation mod 3^3 (instead of 3^4), then it works. There are several instances where using Tq_eigenvalue requires dropping some p-adic or w-adic precision and the same thing appears to be happening here. Is this a problem with our code or is it systematic? If its the latter case: can we avoid it?

rharron commented 11 years ago

In the meantime, I'm going to rewrite this code to use the call to the pari library and I'll allow it to lower its precision somewhat.

rharron commented 11 years ago

I removed the bug label, since I "fixed" the "bug", but the question remains of whether we can avoid this loss of precision (i.e. whether this is a bug!).

rpollack9974 commented 11 years ago

I think the problem is in this code which decreases precision:

        if val > 0:
            for key in Phi._map._dict.keys():
                Phi._map._dict[key].ordp -= val

Whenever we strict off a power of p like this we also need to reduce the precision to we keeping accordingly.

rpollack9974 commented 11 years ago

Sorry no -- ignore that.

rharron commented 11 years ago

are you back in civilization already?

rpollack9974 commented 11 years ago

I'm back a day early. Rained out...

On Sun, Jun 30, 2013 at 2:12 PM, rharron notifications@github.com wrote:

are you back in civilization already?

— Reply to this email directly or view it on GitHubhttps://github.com/lalitkumarj/OMSCategory/issues/61#issuecomment-20252467 .