Author: tlinnet
Date: Wed Aug 27 13:22:45 2014
New Revision: 25334
URL: http://svn.gna.org/viewcvs/relax?rev=25334&view=rev
Log:
Implemented the use of "Newton" as minimisation algorithm for R2eff curve
fitting instead of simplex.
Running the test script:
test_suite/shared_data/dispersion/Kjaergaard_et_al_2013/2_pre_run_r2eff.py
For 50 Monte-Carlo simulations, the time drop from:
3 minutes and 13 s, to 1 min an 5 seconds.
task #7822(https://gna.org/task/index.php?7822): Implement user function to
estimate R2eff and associated errors for exponential curve fitting.
Modified:
trunk/auto_analyses/relax_disp.py
trunk/specific_analyses/relax_disp/api.py
trunk/specific_analyses/relax_disp/optimisation.py
Modified: trunk/auto_analyses/relax_disp.py
URL:
http://svn.gna.org/viewcvs/relax/trunk/auto_analyses/relax_disp.py?rev=25334&r1=25333&r2=25334&view=diff
==============================================================================
--- trunk/auto_analyses/relax_disp.py (original)
+++ trunk/auto_analyses/relax_disp.py Wed Aug 27 13:22:45 2014
@@ -426,6 +426,10 @@
# The constraints flag.
constraints = False
+ # The minimisation algorithm to use.
+ # Both the Jacobian and Hessian matrix has been specified for
exponential curve-fitting, allowing for the much faster algorithms to be used.
+ min_algor = 'Newton'
+
# Check if all spins contains 'r2eff and it associated error.
has_r2eff = False
@@ -454,10 +458,12 @@
else:
do_minimise = True
constraints = True
+ # The minimisation algorithm to use. If the Jacobian and Hessian
matrix have not been specified for fitting, 'simplex' should be used.
+ min_algor = 'simplex'
# Do the minimisation.
if do_minimise:
- self.interpreter.minimise.execute('simplex',
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=constraints)
+ self.interpreter.minimise.execute(min_algor=min_algor,
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=constraints)
# Model elimination.
if self.eliminate:
@@ -469,6 +475,9 @@
# The constraints flag.
constraints = False
+ # Both the Jacobian and Hessian matrix has been specified for
exponential curve-fitting, allowing for the much faster algorithms to be used.
+ min_algor = 'Newton'
+
# Skip optimisation, if 'r2eff' + 'r2eff_err' is present and
flag for forcing optimisation is not raised.
if has_r2eff and not self.optimise_r2eff:
pass
@@ -482,9 +491,11 @@
do_monte_carlo = True
elif self.mc_sim_all_models or len(self.models) < 2:
+ do_monte_carlo = True
# The constraints flag.
constraints = True
- do_monte_carlo = True
+ # The minimisation algorithm to use. If the Jacobian and Hessian
matrix have not been specified for fitting, 'simplex' should be used.
+ min_algor = 'simplex'
# Do Monte Carlo simulations.
if do_monte_carlo:
@@ -494,7 +505,7 @@
self.interpreter.monte_carlo.setup(number=self.mc_sim_num)
self.interpreter.monte_carlo.create_data()
self.interpreter.monte_carlo.initial_values()
- self.interpreter.minimise.execute('simplex',
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=constraints)
+ self.interpreter.minimise.execute(min_algor=min_algor,
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=constraints)
if self.eliminate:
self.interpreter.eliminate()
self.interpreter.monte_carlo.error_analysis()
Modified: trunk/specific_analyses/relax_disp/api.py
URL:
http://svn.gna.org/viewcvs/relax/trunk/specific_analyses/relax_disp/api.py?rev=25334&r1=25333&r2=25334&view=diff
==============================================================================
--- trunk/specific_analyses/relax_disp/api.py (original)
+++ trunk/specific_analyses/relax_disp/api.py Wed Aug 27 13:22:45 2014
@@ -26,7 +26,7 @@
# Python module imports.
from copy import deepcopy
-from re import search
+from re import match, search
import sys
from types import MethodType
@@ -573,9 +573,44 @@
algor = min_algor
if min_algor == 'Log barrier':
algor = min_options[0]
- allowed = ['grid', 'simplex']
- if algor not in allowed:
- raise RelaxError("Only the 'simplex' minimisation algorithm is
supported for the relaxation dispersion analysis as function gradients are
not implemented.")
+
+ allow = False
+ # Check the model type.
+ if hasattr(cdp, 'model_type'):
+ # Set the model type:
+ model_type = cdp.model_type
+
+ if model_type == MODEL_R2EFF:
+ if match('^[Gg]rid$', algor):
+ allow = True
+
+ elif match('^[Ss]implex$', algor):
+ allow = True
+
+ # Quasi-Newton BFGS minimisation.
+ elif match('^[Bb][Ff][Gg][Ss]$', algor):
+ allow = True
+
+ # Newton minimisation.
+ elif match('^[Nn]ewton$', algor):
+ allow = True
+
+ # If the Jacobian and Hessian matrix have not been specified for
fitting, 'simplex' should be used.
+ else:
+ if match('^[Gg]rid$', algor):
+ allow = True
+
+ elif match('^[Ss]implex$', algor):
+ allow = True
+
+ # Do not allow, if no model has been specified.
+ else:
+ model_type = 'None'
+ # Do not allow.
+ allow = False
+
+ if not allow:
+ raise RelaxError("Minimisation algorithm '%s' is not allowed,
since function gradients for model '%s' is not implemented. Only the
'simplex' minimisation algorithm is supported for the relaxation dispersion
analysis of this model."%(algor, model_type))
# Initialise some empty data pipe structures so that the target
function set up does not fail.
if not hasattr(cdp, 'cpmg_frqs_list'):
Modified: trunk/specific_analyses/relax_disp/optimisation.py
URL:
http://svn.gna.org/viewcvs/relax/trunk/specific_analyses/relax_disp/optimisation.py?rev=25334&r1=25333&r2=25334&view=diff
==============================================================================
--- trunk/specific_analyses/relax_disp/optimisation.py (original)
+++ trunk/specific_analyses/relax_disp/optimisation.py Wed Aug 27 13:22:45
2014
@@ -47,7 +47,12 @@
# C modules.
if C_module_exp_fn:
- from target_functions.relax_fit import setup, func, dfunc, d2func,
back_calc_I
+ from specific_analyses.relax_fit.optimisation import func_wrapper,
dfunc_wrapper, d2func_wrapper
+ from target_functions.relax_fit import setup, back_calc_I
+ # Call the python wrapper function to help with list to numpy array
conversion.
+ func = func_wrapper
+ dfunc = dfunc_wrapper
+ d2func = d2func_wrapper
def back_calc_peak_intensities(spin=None, exp_type=None, frq=None,
offset=None, point=None):
r25334 - in /trunk: auto_analyses/ specific_analyses/relax_disp/