Author: bugman
Date: Wed Jul 23 11:27:38 2008
New Revision: 6941
URL: http://svn.gna.org/viewcvs/relax?rev=6941&view=rev
Log:
Started to add the framework for the addition of gradients and Hessians.
Modified:
branches/rdc_analysis/maths_fns/n_state_model.py
branches/rdc_analysis/specific_fns/n_state_model.py
Modified: branches/rdc_analysis/maths_fns/n_state_model.py
URL:
http://svn.gna.org/viewcvs/relax/branches/rdc_analysis/maths_fns/n_state_model.py?rev=6941&r1=6940&r2=6941&view=diff
==============================================================================
--- branches/rdc_analysis/maths_fns/n_state_model.py (original)
+++ branches/rdc_analysis/maths_fns/n_state_model.py Wed Jul 23 11:27:38 2008
@@ -75,6 +75,11 @@
self.params = 1.0 * init_params # Force a copy of the data to be
stored.
self.total_num_params = len(init_params)
+ # Initialise the function value, gradient, and Hessian.
+ self.chi2 = 0.0
+ self.dchi2 = zeros((self.total_num_params), float64)
+ self.d2chi2 = zeros((self.total_num_params, self.total_num_params),
float64)
+
# Scaling initialisation.
self.scaling_matrix = scaling_matrix
if self.scaling_matrix != None:
@@ -127,6 +132,8 @@
# Store the data.
self.rdcs = rdcs
self.xh_vect = xh_vect
+ self.num_align = len(self.rdcs)
+ self.num_align_params = len(self.rdcs)*5
# RDC errors.
if rdc_errors == None:
@@ -138,8 +145,10 @@
# Back calculated RDC array.
self.rdcs_back_calc = 0.0 * deepcopy(rdcs)
- # Set the target function.
+ # Set the target function, gradient, and Hessian.
self.func = self.func_population
+ self.dfunc = self.dfunc_population
+ self.d2func = self.d2func_population
def func_2domain(self, params):
@@ -214,7 +223,7 @@
simply the SSE normalised to unit variance (the SSE divided by the
error squared).
@param params: The vector of parameter values.
- @type params: list of float
+ @type params: numpy rank-1 array
@return: The chi-squared or SSE value.
@rtype: float
"""
@@ -244,3 +253,77 @@
# Return the chi-squared value.
return chi2_sum
+
+
+ def dfunc_population(self, params):
+ """The gradient function for optimisation of the flexible population
N-state model.
+
+ This function should be passed to the optimisation algorithm. It
accepts, as an array, a
+ vector of parameter values and, using these, returns the chi-squared
gradient corresponding
+ to that coordinate in the parameter space. If no RDC errors are
supplied, then the SSE (the
+ sum of squares error) gradient is returned instead. The chi-squared
gradient is simply the
+ SSE gradient normalised to unit variance (the SSE divided by the
error squared).
+
+ @param params: The vector of parameter values.
+ @type params: numpy rank-1 array
+ @return: The chi-squared or SSE gradient.
+ @rtype: numpy rank-1 array
+ """
+
+ # Scaling.
+ if self.scaling_flag:
+ params = dot(params, self.scaling_matrix)
+
+ # Initial chi-squared (or SSE) gradient.
+ chi2_sum = 0.0
+
+ # Unpack the probabilities (located at the end of the parameter
array).
+ probs = params[-(self.N-1):]
+
+ # Loop over the gradient.
+ for i in xrange(self.total_num_params):
+ # Alignment tensor parameter.
+ if i < self.num_align_params:
+ print "Anm: " + `i` + ", " + `params[i]`
+
+ # Probability parameter.
+ else:
+ print "pc: " + `i` + ", " + `params[i]`
+
+ # Loop over each alignment.
+ for n in xrange(self.num_align):
+ # Extract tensor n from the parameters.
+ tensor_5D = params[5*n:5*n + 5]
+
+ # Loop over the spin systems i.
+ for i in xrange(self.num_spins):
+ # Calculate the average RDC.
+ self.rdcs_back_calc[n, i] = average_rdc_5D(self.xh_vect[i],
self.N, tensor_5D, weights=probs)
+
+ # Calculate and sum the single alignment chi-squared value.
+ chi2_sum = chi2_sum + chi2(self.rdcs[n], self.rdcs_back_calc[n],
self.rdc_errors[n])
+
+ # Diagonal scaling.
+ if self.scaling_flag:
+ self.dchi2 = dot(self.dchi2, self.scaling_matrix)
+
+ # Return a copy of the gradient.
+ return self.dchi2 * 1.0
+
+
+ def d2func_population(self, params):
+ """The Hessian function for optimisation of the flexible population
N-state model.
+
+ This function should be passed to the optimisation algorithm. It
accepts, as an array, a
+ vector of parameter values and, using these, returns the chi-squared
Hessian corresponding
+ to that coordinate in the parameter space. If no RDC errors are
supplied, then the SSE (the
+ sum of squares error) Hessian is returned instead. The chi-squared
Hessian is simply the
+ SSE Hessian normalised to unit variance (the SSE divided by the
error squared).
+
+ @param params: The vector of parameter values.
+ @type params: numpy rank-1 array
+ @return: The chi-squared or SSE Hessian.
+ @rtype: numpy rank-2 array
+ """
+
+
Modified: branches/rdc_analysis/specific_fns/n_state_model.py
URL:
http://svn.gna.org/viewcvs/relax/branches/rdc_analysis/specific_fns/n_state_model.py?rev=6941&r1=6940&r2=6941&view=diff
==============================================================================
--- branches/rdc_analysis/specific_fns/n_state_model.py (original)
+++ branches/rdc_analysis/specific_fns/n_state_model.py Wed Jul 23 11:27:38
2008
@@ -757,9 +757,9 @@
# Minimisation.
if constraints:
- results = generic_minimise(func=model.func, 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=1, print_flag=verbosity)
+ results = generic_minimise(func=model.func, dfunc=model.dfunc,
d2func=model.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=1, print_flag=verbosity)
else:
- results = generic_minimise(func=model.func, args=(),
x0=param_vector, min_algor=min_algor, min_options=min_options,
func_tol=func_tol, grad_tol=grad_tol, maxiter=max_iterations, full_output=1,
print_flag=verbosity)
+ results = generic_minimise(func=model.func, dfunc=model.dfunc,
d2func=model.d2func, args=(), x0=param_vector, min_algor=min_algor,
min_options=min_options, func_tol=func_tol, grad_tol=grad_tol,
maxiter=max_iterations, full_output=1, print_flag=verbosity)
if results == None:
return
r6941 - in /branches/rdc_analysis: maths_fns/n_state_model.py specific_fns/n_state_model.py