Re: r26244 - /trunk/test_suite/system_tests/relax_disp.py -- October 14, 2014 - 12:50

Hi Troels,

If you would like to see if optimising the R20A parameters and dw for
a single spin will fix this (i.e. just the spin-dependent parameters),
a simple way would be to make some temporary changes that you never
commit to the relax repository.  You could hard code the k_AB
parameter directly into the target function to the optimised value you
now have.  And then remove it from the parameter list of the model.
Then turn off the clustering and optimise just this spin (deselecting
all others).  You will quickly see if the Nelder-Mead simplex
algorithm can slide along this curved and narrow tunnel in the R20A+dw
space.

Regards,

Edward


On 10 October 2014 17:05,  <tlinnet@xxxxxxxxxxxxx> wrote:
> Author: tlinnet
> Date: Fri Oct 10 17:05:39 2014
> New Revision: 26244
>
> URL: http://svn.gna.org/viewcvs/relax?rev=26244&view=rev
> Log:
> Created systemtest Relax_disp.test_dx_map_clustered_create_par_file, which 
> must show that relax is not able to find the local minimum under clustered 
> conditions.
>
> When creating the map, the map contain chi2 values, which are lower than 
> the clustered fitted values.
>
> This should not be the case.
>
> I am running a larger map with larger bounds and more increements, which 
> should show that there exist a minimum in the minimisation space with a 
> lower chi2 value.
>
> Bug #22754 (https://gna.org/bugs/index.php?22754): The minimise.calculate() 
> does not calculate chi2 value for clustered residues.
> Task #7860(https://gna.org/task/index.php?7860): When dx_map is issued, 
> create a parameter file which maps parameters to chi2 value.
>
> Modified:
>     trunk/test_suite/system_tests/relax_disp.py
>
> Modified: trunk/test_suite/system_tests/relax_disp.py
> URL: 
> http://svn.gna.org/viewcvs/relax/trunk/test_suite/system_tests/relax_disp.py?rev=26244&r1=26243&r2=26244&view=diff
> ==============================================================================
> --- trunk/test_suite/system_tests/relax_disp.py (original)
> +++ trunk/test_suite/system_tests/relax_disp.py Fri Oct 10 17:05:39 2014
> @@ -3044,7 +3044,7 @@
>          lower = []
>          upper = []
>          for i, param_val in enumerate(dx_point_clustered_min):
> -            param = dx_params[0]
> +            param = dx_params[i]
>              step_val = param_delta * param_val
>              step_length = step_val * dx_inc_sides
>
> @@ -3115,6 +3115,137 @@
>
>          # Assert.
>          self.assertAlmostEqual(test, pre_chi2,  6)
> +
> +
> +    def test_dx_map_clustered_create_par_file(self):
> +        """Test making dx_map for residues under clustered calculation, 
> and the creation of the parameter file.
> +
> +        U{Task #7860<https://gna.org/task/index.php?7860>} : When dx_map 
> is issued, create a parameter file which maps parameters to chi2 value.
> +
> +        This uses the data from paper at 
> U{http://dx.doi.org/10.1073/pnas.0509100103}.  This is CPMG data with a 
> fixed relaxation time period.  Experiment in 0.48 M GuHCl (guanidine 
> hydrochloride).
> +        """
> +
> +        # Define path to data
> +        prev_data_path = status.install_path + 
> sep+'test_suite'+sep+'shared_data'+sep+'dispersion'+sep+'KTeilum_FMPoulsen_MAkke_2006'+sep+'surface_chi2_clustered_fitting'
> +
> +        # Read data.
> +        self.interpreter.results.read(prev_data_path + sep + 
> 'coMDD_-_TSMFK01_-_min_-_32_-_free_spins.bz2')
> +
> +        # Get residue of interest.
> +        cur_spin_id = ":%i@%s"%(65, 'N')
> +        cur_spin_id_str = cur_spin_id .replace('#', '_').replace(':', 
> '_').replace('@', '_')
> +
> +        # Get the spin container.
> +        cur_spin = return_spin(cur_spin_id)
> +
> +        # Get the chi2 value
> +        pre_chi2 = cur_spin.chi2
> +
> +        # Then do a local minimisation.
> +        #self.interpreter.select.spin(":%i@%s"%(2, 'N'))
> +        self.interpreter.minimise.calculate()
> +
> +        # Get the chi2 value after calculation.
> +        calc_chi2 = cur_spin.chi2
> +
> +        # Assert calculation is equal.
> +        self.assertAlmostEqual(pre_chi2, calc_chi2)
> +
> +        # Define dx.map settings.
> +        dx_inc = 2
> +        dx_inc_sides = dx_inc / 2
> +
> +        dx_params = ['dw', 'k_AB', 'r2a']
> +        dx_point_clustered_min = [cur_spin.dw, cur_spin.k_AB, 
> cur_spin.r2a['SQ CPMG - 499.86214000 MHz']]
> +
> +        print("Params for dx map is")
> +        print(dx_params)
> +        print("Point param for dx map is, with chi2=%3.3f"%pre_chi2)
> +        print(dx_point_clustered_min)
> +
> +        # Define file_names.
> +        cur_model = 'TSMFK01'
> +        file_name_map = "%s_map%s" % (cur_model, cur_spin_id_str)
> +        file_name_point = "%s_point%s" % (cur_model, cur_spin_id_str)
> +
> +        # Step-size of parameter is 10 %
> +        param_delta = 0.1
> +
> +        # Determine bounds for lower and upper
> +        #lower = [0.0, 0.0, 5.0]
> +        #upper = [20.0, 6.0, 15.0]
> +
> +        lower = []
> +        upper = []
> +        for i, param_val in enumerate(dx_point_clustered_min):
> +            param = dx_params[i]
> +            step_val = param_delta * param_val
> +            step_length = step_val * dx_inc_sides
> +
> +            # Calculate value
> +            low_val = param_val - step_length
> +            lower.append(low_val)
> +
> +            upp_val = param_val + step_length
> +            upper.append(upp_val)
> +
> +            print("For param %s, lower=%3.3f, upper=%3.3f, 
> step_value=%3.3f, steps=%i, centered at=%3.3f"% (param, low_val, upp_val, 
> step_val, dx_inc, param_val))
> +
> +        # Define temporary folder.
> +        result_dir = self.tmpdir
> +
> +        # For testing.
> +        #result_dir = None
> +        #lower = None
> +        #upper = None
> +        #self.interpreter.relax_disp.cluster(cluster_id='free spins', 
> spin_id=cur_spin_id)
> +
> +        # Then do the map.
> +        self.interpreter.dx.map(params=dx_params, map_type='Iso3D', 
> spin_id=cur_spin_id, inc=dx_inc, lower=lower, upper=upper, axis_incs=10, 
> file_prefix=file_name_map, dir=result_dir, point=dx_point_clustered_min, 
> point_file=file_name_point)
> +
> +        # Print where to locate values.
> +        nr_chi2_val = (dx_inc + 1)**3
> +        print("Nr of chi2 calculations are=%i"%nr_chi2_val)
> +        print("Global chi2=%3.3f, Calc_chi=%3.3f" % (pre_chi2, calc_chi2) )
> +
> +        ## Check for file creation
> +        # Set filepaths.
> +        map_name = get_file_path(file_name=file_name_map, dir=result_dir)
> +        map_cfg = get_file_path(file_name=file_name_map+".cfg", 
> dir=result_dir)
> +        map_net = get_file_path(file_name=file_name_map+".net", 
> dir=result_dir)
> +        map_general = get_file_path(file_name=file_name_map+".general", 
> dir=result_dir)
> +        map_par = get_file_path(file_name=file_name_map+".par", 
> dir=result_dir)
> +
> +        point_general = 
> get_file_path(file_name=file_name_point+".general", dir=result_dir)
> +        point_point = get_file_path(file_name=file_name_point, 
> dir=result_dir)
> +
> +        # Test the files exists.
> +        self.assert_(access(map_cfg, F_OK))
> +        self.assert_(access(map_net, F_OK))
> +        self.assert_(access(map_general, F_OK))
> +        self.assert_(access(map_par, F_OK))
> +        self.assert_(access(point_general, F_OK))
> +        self.assert_(access(point_point, F_OK))
> +
> +        print("\nParams for dx map is")
> +        print(dx_params)
> +        print("Point param for dx map is, with chi2=%3.3f"%pre_chi2)
> +        print(dx_point_clustered_min, "\n")
> +
> +        # Open the parameter chi2 file, and assert the chi2 value in the 
> sorted parameter file is not lower that than the global minimisation.
> +        get_data = extract_data(file=map_par)
> +
> +        # Extract line 1, column 9.
> +        test = float(get_data[1][9])
> +
> +        # Print data if map contain a lower value than the global 
> minimised value.
> +        if test < pre_chi2:
> +            print("\nInitial clustered minimised chi2 value is=%3.3f, 
> whereby the minimum map value is=%3.3f\n" % (pre_chi2, test))
> +            for line in get_data:
> +                print(line)
> +
> +        # Assert that the initial global chi2 is lower than the map value.
> +        self.assert_(pre_chi2 < test)
>
>
>      def test_estimate_r2eff_err(self):
>
>
> _______________________________________________
> 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