Re: bug in fsolve() affecting convergence depending on the returned args

[Christophe Prud'homme]

[ Mercredi 26 Avril 2006 09:02 ]
| So the error happens for N = 32 on the first time through the loop
| over the elements of tt(2:end).  Are you not checking info each time
| fsolve is called?  If not then you could miss a case that is bad and
| only see the last call which might be OK.
|
| OTOH, I think that "iteration is not making good progress" does not
| necessarily mean the solution is really bad.  I think it just means
| that convergence seems to be slow.
John,

I modified the script as you proposed (check info on return for the "good" 
case) but without success
I still have the same behavior(I attached the new scripts)
I now even go up the N=1024 and the "good" case is still good
while the "bad" case fails at N=32 (fsolve sends an exception and the results 
for N> 4 are not available due to failure, so I cannot check if they are 
correct is really bad or not)
Note also that I have tested this on Debian/GNU/Linux : stable/unstable 2.1.73 
2.9.5 and on ia32 and amd64 arch. I get always the same behavior.

here is the log 

N=2 get w and info
N=2 get w and info OK
N=2 get only w
N=2 get only w OK
N=4 get w and info
N=4 get w and info OK
N=4 get only w
N=4 get only w OK
N=8 get w and info
N=8 get w and info OK
N=8 get only w
N=8 get only w OK
N=16 get w and info
N=16 get w and info OK
N=16 get only w
N=16 get only w OK
N=32 get w and info
N=32 get w and info OK
N=32 get only w
ans = error: fsolve: iteration is not making good progress

N=32 get only w Failed
N=64 get w and info
N=64 get w and info OK
N=64 get only w
ans = error: fsolve: iteration is not making good progress

N=64 get only w Failed
N=128 get w and info
N=128 get w and info OK
N=128 get only w
ans = error: fsolve: iteration is not making good progress

N=128 get only w Failed
N=256 get w and info
N=256 get w and info OK
N=256 get only w
ans = error: fsolve: iteration is not making good progress

N=256 get only w Failed
N=512 get w and info
N=512 get w and info OK
N=512 get only w
ans = error: fsolve: iteration is not making good progress

N=512 get only w Failed
N=1024 get w and info
N=1024 get w and info OK
N=1024 get only w
ans = error: fsolve: iteration is not making good progress

N=1024 get only w Failed

attached you have the convergence graphics (good and bad)

note also that, as expected, the convergence rate is 2

octave:7> p=log(abs(e_good(1:end-1)./e_good(2:end)))/log(2); p(1:2:end)
ans =

  -0.98617   1.72502   1.98399   1.99901   1.99994

octave:7> p=log(abs(e_bad(1:end-1)./e_bad(2:end)))/log(2); p(1:2:end)
ans =

  -0.98617   1.72502

-- 
Christophe Prud'homme
EPFL SB IACS CMCS
MA B2 534 (Bâtiment MA)
Station 8
CH-1015 Lausanne
Tel: +41 (0)21 693 25 47
Fax: +41 (0)21 693 43 03

cranknic_bad.m
Description: Text Data

cranknic_test2.m
Description: Text Data

cranknic_good.m
Description: Text Data

cranknic_bad.pdf
Description: Adobe PDF document

cranknic_good.pdf
Description: Adobe PDF document