Re: r25322 - /trunk/specific_analyses/relax_disp/estimate_r2eff.py -- August 27, 2014 - 09:30

Hi,

Again you need the Python wrapper functions in
specific_analyses.relax_fit.optimisation.  Note that func(), dfunc(),
and d2func() are the chi-squared value, gradient, and Hessian.  The
Jacobian is a completely different matrix that you can obtain using
the target_functions.relax_fit.jacobian() function, which returns a
Python list of lists, and that only the Levenberg-Marquardt
optimisation algorithm in minfx accepts this.

Regards,

Edward


On 27 August 2014 09:19,  <tlinnet@xxxxxxxxxxxxx> wrote:
> Author: tlinnet
> Date: Wed Aug 27 09:19:45 2014
> New Revision: 25322
>
> URL: http://svn.gna.org/viewcvs/relax?rev=25322&view=rev
> Log:
> Modified estimate R2eff module, to use C-code.
>
> But systemtest Relax_disp.test_estimate_r2eff_error shows that the Jacobian 
> is not correctly implemented
> to be called in minfx.
>
> Modified:
>     trunk/specific_analyses/relax_disp/estimate_r2eff.py
>
> Modified: trunk/specific_analyses/relax_disp/estimate_r2eff.py
> URL: 
> http://svn.gna.org/viewcvs/relax/trunk/specific_analyses/relax_disp/estimate_r2eff.py?rev=25322&r1=25321&r2=25322&view=diff
> ==============================================================================
> --- trunk/specific_analyses/relax_disp/estimate_r2eff.py        (original)
> +++ trunk/specific_analyses/relax_disp/estimate_r2eff.py        Wed Aug 27 
> 09:19:45 2014
> @@ -41,6 +41,8 @@
>  from specific_analyses.relax_disp.parameters import 
> disassemble_param_vector
>  from specific_analyses.relax_disp.variables import MODEL_R2EFF
>  from target_functions.chi2 import chi2_rankN
> +from target_functions.relax_fit import setup, func, dfunc, d2func, 
> back_calc_I
> +
>
>  # Scipy installed.
>  if scipy_module:
> @@ -381,7 +383,7 @@
>
>  # 'minfx'
>  # 'scipy.optimize.leastsq'
> -def estimate_r2eff(spin_id=None, ftol=1e-15, xtol=1e-15, maxfev=10000000, 
> factor=100.0, method='scipy.optimize.leastsq', verbosity=1):
> +def estimate_r2eff(spin_id=None, ftol=1e-15, xtol=1e-15, maxfev=10000000, 
> factor=100.0, method='minfx', verbosity=1):
>      """Estimate r2eff and errors by exponential curve fitting with 
> scipy.optimize.leastsq.
>
>      scipy.optimize.leastsq is a wrapper around MINPACK's lmdif and lmder 
> algorithms.
> @@ -703,19 +705,33 @@
>
>      E.set_settings_minfx(min_algor=min_algor)
>
> -    # Define function to minimise for minfx.
> -    if match('^[Ss]implex$', E.min_algor):
> -        func = E.func_exp
> -
> -        dfunc = None
> -        d2func = None
> +    # Do C code
> +    do_C = True
> +
> +    if do_C:
> +        # Initialise the function to minimise.
> +        scaling_list = [1.0, 1.0]
> +        setup(num_params=len(x0), num_times=len(E.times), values=E.values, 
> sd=E.errors, relax_times=E.times, scaling_matrix=scaling_list)
> +
> +        t_func = func
> +        t_dfunc = dfunc
> +        t_d2func = d2func
> +
>      else:
> -        func = E.func_exp
> -        dfunc = E.func_exp_grad
> -        d2func = E.func_exp_hess
> -
> -    # Minimise with minfx.
> -    results_minfx = generic_minimise(func=func, dfunc=dfunc, 
> d2func=d2func, args=(), x0=x0, min_algor=E.min_algor, 
> min_options=E.min_options, func_tol=E.func_tol, grad_tol=E.grad_tol, 
> maxiter=E.max_iterations, A=E.A, b=E.b, full_output=True, 
> print_flag=E.verbosity)
> +        # Minimise with minfx.
> +        # Define function to minimise for minfx.
> +        if match('^[Ss]implex$', E.min_algor):
> +            t_func = E.func_exp
> +
> +            t_dfunc = None
> +            t_d2func = None
> +        else:
> +            t_func = E.func_exp
> +            t_dfunc = E.func_exp_grad
> +            t_d2func = E.func_exp_hess
> +
> +    # Minimise.
> +    results_minfx = generic_minimise(func=t_func, dfunc=t_dfunc, 
> d2func=t_d2func, args=(), x0=x0, min_algor=E.min_algor, 
> min_options=E.min_options, func_tol=E.func_tol, grad_tol=E.grad_tol, 
> maxiter=E.max_iterations, A=E.A, b=E.b, full_output=True, 
> print_flag=E.verbosity)
>
>      # Unpack
>      param_vector, chi2, iter_count, f_count, g_count, h_count, warning = 
> results_minfx
>
>
> _______________________________________________
> 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