Changes in src/sans/fit/ParkFitting.py [8d074d9:9d6d5ba] in sasview
- File:
-
- 1 edited
-
src/sans/fit/ParkFitting.py (modified) (9 diffs)
Legend:
- Unmodified
- Added
- Removed
-
src/sans/fit/ParkFitting.py
r8d074d9 r9d6d5ba 24 24 from sans.fit.AbstractFitEngine import FitEngine 25 25 from sans.fit.AbstractFitEngine import FResult 26 27 class SansParameter(park.Parameter): 28 """ 29 SANS model parameters for use in the PARK fitting service. 30 The parameter attribute value is redirected to the underlying 31 parameter value in the SANS model. 32 """ 33 def __init__(self, name, model, data): 34 """ 35 :param name: the name of the model parameter 36 :param model: the sans model to wrap as a park model 37 """ 38 park.Parameter.__init__(self, name) 39 #self._model, self._name = model, name 40 self.data = data 41 self.model = model 42 #set the value for the parameter of the given name 43 self.set(model.getParam(name)) 44 45 # TODO: model is missing parameter ranges for dispersion parameters 46 if name not in model.details: 47 #print "setting details for",name 48 model.details[name] = ["", None, None] 49 50 def _getvalue(self): 51 """ 52 override the _getvalue of park parameter 53 54 :return value the parameter associates with self.name 55 56 """ 57 return self.model.getParam(self.name) 58 59 def _setvalue(self, value): 60 """ 61 override the _setvalue pf park parameter 62 63 :param value: the value to set on a given parameter 64 65 """ 66 self.model.setParam(self.name, value) 67 68 value = property(_getvalue, _setvalue) 69 70 def _getrange(self): 71 """ 72 Override _getrange of park parameter 73 return the range of parameter 74 """ 75 #if not self.name in self._model.getDispParamList(): 76 lo, hi = self.model.details[self.name][1:3] 77 if lo is None: lo = -numpy.inf 78 if hi is None: hi = numpy.inf 79 if lo > hi: 80 raise ValueError, "wrong fit range for parameters" 81 82 return lo, hi 83 84 def get_name(self): 85 """ 86 """ 87 return self._getname() 88 89 def _setrange(self, r): 90 """ 91 override _setrange of park parameter 92 93 :param r: the value of the range to set 94 95 """ 96 self.model.details[self.name][1:3] = r 97 range = property(_getrange, _setrange) 98 99 100 class ParkModel(park.Model): 101 """ 102 PARK wrapper for SANS models. 103 """ 104 def __init__(self, sans_model, sans_data=None, **kw): 105 """ 106 :param sans_model: the sans model to wrap using park interface 107 108 """ 109 park.Model.__init__(self, **kw) 110 self.model = sans_model 111 self.name = sans_model.name 112 self.data = sans_data 113 #list of parameters names 114 self.sansp = sans_model.getParamList() 115 #list of park parameter 116 self.parkp = [SansParameter(p, sans_model, sans_data) for p in self.sansp] 117 #list of parameter set 118 self.parameterset = park.ParameterSet(sans_model.name, pars=self.parkp) 119 self.pars = [] 120 121 def get_params(self, fitparams): 122 """ 123 return a list of value of paramter to fit 124 125 :param fitparams: list of paramaters name to fit 126 127 """ 128 list_params = [] 129 self.pars = fitparams 130 for item in fitparams: 131 for element in self.parkp: 132 if element.name == str(item): 133 list_params.append(element.value) 134 return list_params 135 136 def set_params(self, paramlist, params): 137 """ 138 Set value for parameters to fit 139 140 :param params: list of value for parameters to fit 141 142 """ 143 try: 144 for i in range(len(self.parkp)): 145 for j in range(len(paramlist)): 146 if self.parkp[i].name == paramlist[j]: 147 self.parkp[i].value = params[j] 148 self.model.setParam(self.parkp[i].name, params[j]) 149 except: 150 raise 151 152 def eval(self, x): 153 """ 154 Override eval method of park model. 155 156 :param x: the x value used to compute a function 157 """ 158 try: 159 return self.model.evalDistribution(x) 160 except: 161 raise 162 163 def eval_derivs(self, x, pars=[]): 164 """ 165 Evaluate the model and derivatives wrt pars at x. 166 167 pars is a list of the names of the parameters for which derivatives 168 are desired. 169 170 This method needs to be specialized in the model to evaluate the 171 model function. Alternatively, the model can implement is own 172 version of residuals which calculates the residuals directly 173 instead of calling eval. 174 """ 175 return [] 176 177 26 178 27 class SansFitResult(fitresult.FitResult): 179 28 def __init__(self, *args, **kwrds): … … 395 244 return fitpars 396 245 397 def extend_results_with_calculated_parameters(self, result):246 def all_results(self, result): 398 247 """ 399 248 Extend result from the fit with the calculated parameters. … … 443 292 # dividing residuals by N in order to be consistent with Scipy 444 293 m.chisq = numpy.sum(m.residuals**2/N) 445 resid.append(m.weight*m.residuals )294 resid.append(m.weight*m.residuals/math.sqrt(N)) 446 295 self.residuals = numpy.hstack(resid) 447 296 N = len(self.residuals) 448 297 self.degrees_of_freedom = N-k if N>k else 1 449 298 self.chisq = numpy.sum(self.residuals**2) 450 return self.chisq /self.degrees_of_freedom299 return self.chisq 451 300 452 301 class ParkFit(FitEngine): … … 505 354 if fproblem.get_to_fit() == 1: 506 355 fitproblems.append(fproblem) 507 if len(fitproblems) == 0: 356 if len(fitproblems) == 0: 508 357 raise RuntimeError, "No Assembly scheduled for Park fitting." 358 return 509 359 for item in fitproblems: 510 model = item.get_model() 511 parkmodel = ParkModel(model.model, model.data) 512 parkmodel.pars = item.pars 360 parkmodel = item.get_model() 513 361 if reset_flag: 514 362 # reset the initial value; useful for batch … … 516 364 ind = item.pars.index(name) 517 365 parkmodel.model.setParam(name, item.vals[ind]) 518 519 # set the constraints into the model520 for p,v in item.constraints:521 parkmodel.parameterset[str(p)].set(str(v))522 366 523 367 for p in parkmodel.parameterset: 524 368 ## does not allow status change for constraint parameters 525 369 if p.status != 'computed': 526 if p.get_name() 370 if p.get_name()in item.pars: 527 371 ## make parameters selected for 528 372 #fit will be between boundaries … … 539 383 def fit(self, msg_q=None, 540 384 q=None, handler=None, curr_thread=None, 541 ftol=1.49012e-8, reset_flag=False):385 ftol=1.49012e-8, reset_flag=False): 542 386 """ 543 387 Performs fit with park.fit module.It can perform fit with one model … … 563 407 localfit = SansFitSimplex() 564 408 localfit.ftol = ftol 565 localfit.xtol = 1e-6 566 409 567 410 # See `park.fitresult.FitHandler` for details. 568 411 fitter = SansFitMC(localfit=localfit, start_points=1) … … 573 416 try: 574 417 result = fit.fit(self.problem, fitter=fitter, handler=handler) 575 self.problem. extend_results_with_calculated_parameters(result)418 self.problem.all_results(result) 576 419 577 420 except LinAlgError: 578 421 raise ValueError, "SVD did not converge" 579 580 if result is None:581 raise RuntimeError("park did not return a fit result")582 422 583 423 for m in self.problem.parts: … … 587 427 small_result.theory = theory 588 428 small_result.residuals = residuals 589 small_result.index = m.data.idx 590 small_result.fitness = result.fitness 591 592 # Extract the parameters that are part of this model; make sure 593 # they match the fitted parameters for this model, and place them 594 # in the same order as they occur in the model. 595 pars = {} 596 for p in result.parameters: 597 #if p.data.name == small_result.data.name and 598 if p.model.name == small_result.model.name: 599 model_name, par_name = p.name.split('.', 1) 600 pars[par_name] = (p.value, p.stderr) 601 #assert len(pars.keys()) == len(m.model.pars) 602 v,dv = zip(*[pars[p] for p in m.model.pars]) 603 small_result.pvec = v 604 small_result.stderr = dv 605 small_result.param_list = m.model.pars 606 607 # normalize chisq by degrees of freedom 608 dof = len(small_result.residuals)-len(small_result.pvec) 609 small_result.fitness = numpy.sum(residuals**2)/dof 610 429 small_result.pvec = [] 430 small_result.cov = [] 431 small_result.stderr = [] 432 small_result.param_list = [] 433 small_result.residuals = m.residuals 434 if result is not None: 435 for p in result.parameters: 436 if p.data.name == small_result.data.name and \ 437 p.model.name == small_result.model.name: 438 small_result.index = m.data.idx 439 small_result.fitness = result.fitness 440 small_result.pvec.append(p.value) 441 small_result.stderr.append(p.stderr) 442 name_split = p.name.split('.') 443 name = name_split[1].strip() 444 if len(name_split) > 2: 445 name += '.' + name_split[2].strip() 446 small_result.param_list.append(name) 611 447 result_list.append(small_result) 612 448 if q != None:
Note: See TracChangeset
for help on using the changeset viewer.
