Re: r23277 - in /branches/disp_speed: lib/dispersion/m61.py target_functions/relax_disp.py -- May 21, 2014

Hi Edward.

Thanks for all the suggestions for correcting the error cathing.

I am looking into it, and I am very happy that you catch my errors.

The reason why I did not split up the commit in two, as you have
suggested several times now,
was that I would not have been able to do the return command.

          But that was rubbish thinking!!!

I could have implemented the return command first for all models, and
then afterwards do all the error catching.
That could have made it easy for myself to revert the commits which
only deals with the error catching.
Sigh...., my logic thinking suffers from late night working. And my
stubbornness have here back-fired.

And I am also sorry for the trouble this causes, when you have to
review all the commits to be put into the announcement with the list
of code changes.
One can probably not fathom this amount of work, and cursing, before
one have done itself. :-)

Allright.

Line-up of work.
1) I will compare with trunk, and I will only implement the error
cathing which was as before.

2) I will add:
# Catch parameter values that will result in no exchange, returning
flat R2eff = R20 lines.
if dw == 0.0 or pA == 1.0 or kex == 0.0:
    return array([R20]*num_points)

3) I will keep this check, as a "last line of defense".
    # Catch errors, taking a sum over array is the fastest way to check for
    # +/- inf (infinity) and nan (not a number).
    if not isfinite(sum(R2eff)):
        R2eff = array([1e100]*num_points)


I hope you can bear with me, and I will again try to switch on the brain. :-)

Best
Troels





2014-05-21 9:00 GMT+02:00 Edward d'Auvergne <edward@xxxxxxxxxxxxx>:

Hi,

Again I don't think that the denominator can be zero.  Did you see
this in the test suite?

Regards,

Edward



On 20 May 2014 22:29,  <tlinnet@xxxxxxxxxxxxx> wrote:

Author: tlinnet
Date: Tue May 20 22:29:52 2014
New Revision: 23277

URL: http://svn.gna.org/viewcvs/relax?rev=23277&view=rev
Log:
Math-domain catching for model M61.

task #7793: (https://gna.org/task/?7793) Speed-up of dispersion models.

This is to implement catching of math domain errors, before they occur.
These can be found via the --numpy-raise function to the systemtests.

To make the code look clean, the class object "back_calc" is no longer
being updated per time point, but is updated in the relax_disp target 
function in
one go.

Modified:
    branches/disp_speed/lib/dispersion/m61.py
    branches/disp_speed/target_functions/relax_disp.py

Modified: branches/disp_speed/lib/dispersion/m61.py
URL: 
http://svn.gna.org/viewcvs/relax/branches/disp_speed/lib/dispersion/m61.py?rev=23277&r1=23276&r2=23277&view=diff
==============================================================================
--- branches/disp_speed/lib/dispersion/m61.py   (original)
+++ branches/disp_speed/lib/dispersion/m61.py   Tue May 20 22:29:52 2014
@@ -64,10 +64,10 @@
 """

 # Python module imports.
-from numpy import array, isfinite, sum
+from numpy import abs, array, isfinite, min, sum


-def r1rho_M61(r1rho_prime=None, phi_ex=None, kex=None, 
spin_lock_fields2=None, back_calc=None, num_points=None):
+def r1rho_M61(r1rho_prime=None, phi_ex=None, kex=None, 
spin_lock_fields2=None, num_points=None):
     """Calculate the R2eff values for the M61 model.

     See the module docstring for details.
@@ -81,9 +81,7 @@
     @type kex:                  float
     @keyword spin_lock_fields2: The R1rho spin-lock field strengths 
squared (in rad^2.s^-2).
     @type spin_lock_fields2:    numpy rank-1 float array
-    @keyword back_calc:         The array for holding the back calculated 
R1rho values.  Each element corresponds to one of the spin-lock fields.
-    @type back_calc:            numpy rank-1 float array
-    @keyword num_points:        The number of points on the dispersion 
curve, equal to the length of the spin_lock_fields and back_calc arguments.
+    @keyword num_points:        The number of points on the dispersion 
curve, equal to the length of the spin_lock_fields.
     @type num_points:           int
     """

@@ -96,6 +94,13 @@
     # Denominator.
     denom = kex2 + spin_lock_fields2

+    # Catch math domain error of dividing with 0.
+    # This is when denom=0.
+    if min(abs(denom)) == 0:
+        R1rho = array([1e100]*num_points)
+
+        return R1rho
+
     # R1rho calculation.
     R1rho = r1rho_prime + numer / denom

@@ -104,6 +109,4 @@
     if not isfinite(sum(R1rho)):
         R1rho = array([1e100]*num_points)

-    # Parse back the value to update the back_calc class object.
-    for i in range(num_points):
-        back_calc[i] = R1rho[i]
+    return R1rho

Modified: branches/disp_speed/target_functions/relax_disp.py
URL: 
http://svn.gna.org/viewcvs/relax/branches/disp_speed/target_functions/relax_disp.py?rev=23277&r1=23276&r2=23277&view=diff
==============================================================================
--- branches/disp_speed/target_functions/relax_disp.py  (original)
+++ branches/disp_speed/target_functions/relax_disp.py  Tue May 20 
22:29:52 2014
@@ -1163,7 +1163,7 @@
                 phi_ex_scaled = phi_ex[si] * self.frqs[0][si][mi]**2

                 # Back calculate the R2eff values.
-                r1rho_M61(r1rho_prime=R20[r20_index], 
phi_ex=phi_ex_scaled, kex=kex, 
spin_lock_fields2=self.spin_lock_omega1_squared[0][mi][0], 
back_calc=self.back_calc[0][si][mi][0], 
num_points=self.num_disp_points[0][si][mi][0])
+                self.back_calc[0][si][mi][0] = 
r1rho_M61(r1rho_prime=R20[r20_index], phi_ex=phi_ex_scaled, kex=kex, 
spin_lock_fields2=self.spin_lock_omega1_squared[0][mi][0], 
num_points=self.num_disp_points[0][si][mi][0])

                 # For all missing data points, set the back-calculated 
value to the measured values so that it has no effect on the chi-squared 
value.
                 for di in range(self.num_disp_points[0][si][mi][0]):


_______________________________________________
relax (http://www.nmr-relax.com)

This is the relax-commits mailing list
relax-commits@xxxxxxx

To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-commits


_______________________________________________
relax (http://www.nmr-relax.com)

This is the relax-devel mailing list
relax-devel@xxxxxxx

To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-devel

Re: r23277 - in /branches/disp_speed: lib/dispersion/m61.py target_functions/relax_disp.py

Header

Content

Related Messages