Oh, right, that's another 'feature' I didn't implement as I was too busy finishing things off to publish our dispersion analysis in relax paper :S It also would have served no purpose - if there are gradients implemented for a model, then we can use the Methods-of-Multipliers as a better alternative algorithm anyway.
Cheers, Edward On 1 September 2014 13:08, Troels Emtekær Linnet <[email protected]> wrote: > Hi Edward. > > If I in systemtest: > test_estimate_r2eff_err > > change: > self.interpreter.minimise.execute(min_algor='Newton', > constraints=False, verbosity=1) > to > self.interpreter.minimise.execute(min_algor='Newton', > constraints=True, verbosity=1) > > Then I get: > > ############################################## > > relax> minimise.grid_search(lower=None, upper=None, inc=11, > verbosity=1, constraints=True, skip_preset=True) > > > Grid search setup: the spin block [':52@N'] > -------------------------------------------- > > ...... > > relax> minimise.execute(min_algor='Newton', line_search=None, > hessian_mod=None, hessian_type=None, func_tol=1e-25, grad_tol=None, > max_iter=10000000, constraints=True, scaling=True, verbosity=1) > Resetting the minimisation statistics. > > > Fitting to spin :52@N, frequency 799777399.1 and dispersion point 431.0 > ----------------------------------------------------------------------- > > > > Logarithmic barrier function > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > k: 0 xk: [ 8.8, 200000.0001] fk: 37.0428718161 > Entering sub-algorithm. > > Newton minimisation > ~~~~~~~~~~~~~~~~~~~ > Line search: Backtracking line search. > Hessian modification: The Gill, Murray, and Wright modified > Cholesky algorithm. > E > ====================================================================== > ERROR: test_estimate_r2eff_err (test_suite.system_tests.relax_disp.Relax_disp) > Test the user function for estimating R2eff errors from exponential > curve fitting. > ---------------------------------------------------------------------- > Traceback (most recent call last): > File > "/sbinlab2/tlinnet/software/NMR-relax/relax_trunk/test_suite/system_tests/relax_disp.py", > line 2990, in test_estimate_r2eff_err > self.interpreter.minimise.execute(min_algor='Newton', > constraints=True, verbosity=1) > File > "/sbinlab2/tlinnet/software/NMR-relax/relax_trunk/prompt/uf_objects.py", > line 223, in __call__ > self._backend(*new_args, **uf_kargs) > File > "/sbinlab2/tlinnet/software/NMR-relax/relax_trunk/pipe_control/minimise.py", > line 527, in minimise > api.minimise(min_algor=min_algor, min_options=min_options, > func_tol=func_tol, grad_tol=grad_tol, max_iterations=max_iter, > constraints=constraints, scaling_matrix=scaling_matrix, > verbosity=verbosity) > File > "/sbinlab2/tlinnet/software/NMR-relax/relax_trunk/specific_analyses/relax_disp/api.py", > line 668, in minimise > minimise_r2eff(spins=spins, spin_ids=spin_ids, > min_algor=min_algor, min_options=min_options, func_tol=func_tol, > grad_tol=grad_tol, max_iterations=max_iterations, > constraints=constraints, scaling_matrix=scaling_matrix[model_index], > verbosity=verbosity, sim_index=sim_index, lower=lower_i, > upper=upper_i, inc=inc_i) > File > "/sbinlab2/tlinnet/software/NMR-relax/relax_trunk/specific_analyses/relax_disp/optimisation.py", > line 424, in minimise_r2eff > results = generic_minimise(func=func, dfunc=dfunc, d2func=d2func, > args=(), x0=param_vector, min_algor=min_algor, > min_options=min_options, func_tol=func_tol, grad_tol=grad_tol, > maxiter=max_iterations, A=A, b=b, full_output=True, > print_flag=verbosity) > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/generic.py", > line 399, in generic_minimise > results = log_barrier_function(func=func, dfunc=dfunc, > d2func=d2func, args=args, x0=x0, min_options=min_options, A=A, b=b, > func_tol=func_tol, grad_tol=grad_tol, maxiter=maxiter, > full_output=full_output, print_flag=print_flag) > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/log_barrier_function.py", > line 96, in log_barrier_function > results = min.minimise() > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/log_barrier_function.py", > line 264, in minimise > results = self.generic_minimise(func=self.func_log, > dfunc=self.func_dlog, d2func=self.func_d2log, args=self.args, > x0=self.xk, min_algor=self.min_algor, min_options=self.min_options, > func_tol=self.func_tol, grad_tol=self.grad_tol, maxiter=maxiter, > full_output=1, print_flag=self.print_flag, print_prefix="\t") > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/generic.py", > line 326, in generic_minimise > results = newton(func=func, dfunc=dfunc, d2func=d2func, args=args, > x0=x0, min_options=min_options, func_tol=func_tol, grad_tol=grad_tol, > maxiter=maxiter, full_output=full_output, print_flag=print_flag, > print_prefix=print_prefix) > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/newton.py", > line 47, in newton > min = Newton(func, dfunc, d2func, args, x0, min_options, func_tol, > grad_tol, maxiter, a0, mu, eta, mach_acc, full_output, print_flag, > print_prefix) > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/newton.py", > line 156, in __init__ > self.setup_newton() > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/newton.py", > line 211, in setup_newton > self.dfk, self.g_count = self.dfunc(*(self.xk,)+self.args), self.g_count > + 1 > File > "/sbinlab2/software/python-enthought-dis/canopy-1.4.0-full-rh5-64/Canopy_64bit/User/lib/python2.7/site-packages/minfx/log_barrier_function.py", > line 211, in func_dlog > raise NameError("The logarithmic barrier gradient is not implemented > yet.") > NameError: The logarithmic barrier gradient is not implemented yet. > > ---------------------------------------------------------------------- > > > > 2014-09-01 12:42 GMT+02:00 Edward d'Auvergne <[email protected]>: >> On 1 September 2014 12:34, Troels Emtekær Linnet <[email protected]> >> wrote: >>> Anyway, before minfx can handle constraints in for example BFGS, >>> this is just a waste of time. >> >> Minfx can do this :) The log-barrier constraint algorithm works with >> all optimisation techniques in minfx, well, apart from the grid search >> (https://en.wikipedia.org/wiki/Barrier_function#Logarithmic_barrier_function). >> And if gradients are supplied, the more powerful >> Methods-of-Multipliers algorithm can also be used in combination with >> all optimisation techniques >> (https://en.wikipedia.org/wiki/Augmented_Lagrangian_method). >> >> >>> I think there will be a 10 x speed up, just for the Jacobian. >> >> For the analytic models, you could have a 10x speed up if symbolic >> gradients and Hessians are implemented. I'm guessing that's what you >> mean. >> >> >>> And when you have the Jacobian, estimating the errors are trivial. >>> >>> std(q) = sqrt ( (dq/dx std(x))*2 + (dq/dz std(z))*2 ) >> >> :S I'm not sure about this estimate. It looks rather too linear. I >> wish errors would be so simple. >> >> >>> where q is the function. x and z are R1 and R1rho_prime. >>> >>> So, until then, implementing the Jacobian is only for testing the >>> error estimation compared to >>> Monte-Carlo simulations. >> >> If you do add the equations, the lib.dispersion.dpl94 module would be >> the natural place to put them. And the interface as dfunc_DPL94(), >> d2func_DPL94(), and jacobian_DPL94(). >> >> Regards, >> >> Edward _______________________________________________ relax (http://www.nmr-relax.com) This is the relax-devel mailing list [email protected] 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
