[aa36f96] | 1 | |
---|
| 2 | |
---|
| 3 | |
---|
[792db7d5] | 4 | """ |
---|
[aa36f96] | 5 | ParkFitting module contains SansParameter,Model,Data |
---|
[b2f25dc5] | 6 | FitArrange, ParkFit,Parameter classes.All listed classes work together |
---|
| 7 | to perform a simple fit with park optimizer. |
---|
[792db7d5] | 8 | """ |
---|
[89f3b66] | 9 | #import time |
---|
[dfd3577] | 10 | import numpy |
---|
[b2f25dc5] | 11 | #import park |
---|
| 12 | from park import fit |
---|
| 13 | from park import fitresult |
---|
[1cff677] | 14 | from park.fitresult import FitParameter |
---|
| 15 | import park.simplex |
---|
[b2f25dc5] | 16 | from park.assembly import Assembly |
---|
[1cff677] | 17 | from park.assembly import Part |
---|
[b2f25dc5] | 18 | from park.fitmc import FitSimplex |
---|
[dfd3577] | 19 | import park.fitmc |
---|
[b2f25dc5] | 20 | from park.fitmc import FitMC |
---|
[1cff677] | 21 | from park.fit import Fitter |
---|
| 22 | from park.formatnum import format_uncertainty |
---|
[61cb28d] | 23 | #from Loader import Load |
---|
[b2f25dc5] | 24 | from sans.fit.AbstractFitEngine import FitEngine |
---|
[1cff677] | 25 | |
---|
[634ca14] | 26 | class SansFitResult(fitresult.FitResult): |
---|
| 27 | def __init__(self, *args, **kwrds): |
---|
| 28 | fitresult.FitResult.__init__(self, *args, **kwrds) |
---|
[425e49ca] | 29 | self.theory = None |
---|
[634ca14] | 30 | self.inputs = [] |
---|
| 31 | |
---|
[1cff677] | 32 | class SansFitSimplex(FitSimplex): |
---|
| 33 | """ |
---|
| 34 | Local minimizer using Nelder-Mead simplex algorithm. |
---|
| 35 | |
---|
| 36 | Simplex is robust and derivative free, though not very efficient. |
---|
| 37 | |
---|
| 38 | This class wraps the bounds contrained Nelder-Mead simplex |
---|
| 39 | implementation for `park.simplex.simplex`. |
---|
| 40 | """ |
---|
| 41 | radius = 0.05 |
---|
| 42 | """Size of the initial simplex; this is a portion between 0 and 1""" |
---|
| 43 | xtol = 1 |
---|
| 44 | #xtol = 1e-4 |
---|
| 45 | """Stop when simplex vertices are within xtol of each other""" |
---|
[425e49ca] | 46 | ftol = 5e-5 |
---|
[1cff677] | 47 | """Stop when vertex values are within ftol of each other""" |
---|
| 48 | maxiter = None |
---|
| 49 | """Maximum number of iterations before fit terminates""" |
---|
| 50 | def fit(self, fitness, x0): |
---|
| 51 | """Run the fit""" |
---|
| 52 | self.cancel = False |
---|
| 53 | pars = fitness.fit_parameters() |
---|
| 54 | bounds = numpy.array([p.range for p in pars]).T |
---|
| 55 | result = park.simplex.simplex(fitness, x0, bounds=bounds, |
---|
| 56 | radius=self.radius, xtol=self.xtol, |
---|
| 57 | ftol=self.ftol, maxiter=self.maxiter, |
---|
| 58 | abort_test=self._iscancelled) |
---|
| 59 | #print "calls:",result.calls |
---|
| 60 | #print "simplex returned",result.x,result.fx |
---|
| 61 | # Need to make our own copy of the fit results so that the |
---|
| 62 | # values don't get stomped on by the next fit iteration. |
---|
| 63 | fitpars = [SansFitParameter(pars[i].name,pars[i].range,v, pars[i].model, pars[i].data) |
---|
| 64 | for i,v in enumerate(result.x)] |
---|
[634ca14] | 65 | res = SansFitResult(fitpars, result.calls, result.fx) |
---|
| 66 | res.inputs = [(pars[i].model, pars[i].data) for i,v in enumerate(result.x)] |
---|
[1cff677] | 67 | # Compute the parameter uncertainties from the jacobian |
---|
| 68 | res.calc_cov(fitness) |
---|
| 69 | return res |
---|
| 70 | |
---|
| 71 | class SansFitter(Fitter): |
---|
| 72 | """ |
---|
| 73 | """ |
---|
| 74 | def fit(self, fitness, handler): |
---|
| 75 | """ |
---|
| 76 | Global optimizer. |
---|
| 77 | |
---|
| 78 | This function should return immediately |
---|
| 79 | """ |
---|
| 80 | # Determine initial value and bounds |
---|
| 81 | pars = fitness.fit_parameters() |
---|
| 82 | bounds = numpy.array([p.range for p in pars]).T |
---|
| 83 | x0 = [p.value for p in pars] |
---|
| 84 | |
---|
| 85 | # Initialize the monitor and results. |
---|
| 86 | # Need to make our own copy of the fit results so that the |
---|
| 87 | # values don't get stomped on by the next fit iteration. |
---|
| 88 | handler.done = False |
---|
| 89 | self.handler = handler |
---|
| 90 | fitpars = [SansFitParameter(pars[i].name, pars[i].range, v, |
---|
| 91 | pars[i].model, pars[i].data) |
---|
| 92 | for i,v in enumerate(x0)] |
---|
| 93 | handler.result = fitresult.FitResult(fitpars, 0, numpy.NaN) |
---|
[d4b0687] | 94 | |
---|
[1cff677] | 95 | # Run the fit (fit should perform _progress and _improvement updates) |
---|
| 96 | # This function may return before the fit is complete. |
---|
| 97 | self._fit(fitness, x0, bounds) |
---|
| 98 | |
---|
| 99 | class SansFitMC(SansFitter): |
---|
| 100 | """ |
---|
| 101 | Monte Carlo optimizer. |
---|
| 102 | |
---|
| 103 | This implements `park.fit.Fitter`. |
---|
| 104 | """ |
---|
| 105 | localfit = SansFitSimplex() |
---|
| 106 | start_points = 10 |
---|
| 107 | |
---|
| 108 | def _fit(self, objective, x0, bounds): |
---|
| 109 | """ |
---|
| 110 | Run a monte carlo fit. |
---|
| 111 | |
---|
| 112 | This procedure maps a local optimizer across a set of initial points. |
---|
| 113 | """ |
---|
| 114 | park.fitmc.fitmc(objective, x0, bounds, self.localfit, |
---|
| 115 | self.start_points, self.handler) |
---|
| 116 | |
---|
| 117 | |
---|
| 118 | class SansPart(Part): |
---|
| 119 | """ |
---|
| 120 | Part of a fitting assembly. Part holds the model itself and |
---|
| 121 | associated data. The part can be initialized with a fitness |
---|
| 122 | object or with a pair (model,data) for the default fitness function. |
---|
[fadea71] | 123 | |
---|
[1cff677] | 124 | fitness (Fitness) |
---|
| 125 | object implementing the `park.assembly.Fitness` interface. In |
---|
| 126 | particular, fitness should provide a parameterset attribute |
---|
| 127 | containing a ParameterSet and a residuals method returning a vector |
---|
| 128 | of residuals. |
---|
| 129 | weight (dimensionless) |
---|
| 130 | weight for the model. See comments in assembly.py for details. |
---|
| 131 | isfitted (boolean) |
---|
| 132 | True if the model residuals should be included in the fit. |
---|
| 133 | The model parameters may still be used in parameter |
---|
| 134 | expressions, but there will be no comparison to the data. |
---|
| 135 | residuals (vector) |
---|
| 136 | Residuals for the model if they have been calculated, or None |
---|
| 137 | degrees_of_freedom |
---|
| 138 | Number of residuals minus number of fitted parameters. |
---|
| 139 | Degrees of freedom for individual models does not make |
---|
| 140 | sense in the presence of expressions combining models, |
---|
| 141 | particularly in the case where a model has many parameters |
---|
| 142 | but no data or many computed parameters. The degrees of |
---|
| 143 | freedom for the model is set to be at least one. |
---|
| 144 | chisq |
---|
| 145 | sum(residuals**2); use chisq/degrees_of_freedom to |
---|
| 146 | get the reduced chisq value. |
---|
[dfd3577] | 147 | |
---|
[1cff677] | 148 | Get/set the weight on the given model. |
---|
| 149 | |
---|
| 150 | assembly.weight(3) returns the weight on model 3 (0-origin) |
---|
| 151 | assembly.weight(3,0.5) sets the weight on model 3 (0-origin) |
---|
| 152 | """ |
---|
| 153 | |
---|
| 154 | def __init__(self, fitness, weight=1., isfitted=True): |
---|
| 155 | Part.__init__(self, fitness=fitness, weight=weight, |
---|
| 156 | isfitted=isfitted) |
---|
| 157 | |
---|
| 158 | self.model, self.data = fitness[0], fitness[1] |
---|
| 159 | |
---|
| 160 | class SansFitParameter(FitParameter): |
---|
| 161 | """ |
---|
| 162 | Fit result for an individual parameter. |
---|
| 163 | """ |
---|
| 164 | def __init__(self, name, range, value, model, data): |
---|
| 165 | FitParameter.__init__(self, name, range, value) |
---|
| 166 | self.model = model |
---|
| 167 | self.data = data |
---|
| 168 | |
---|
| 169 | def summarize(self): |
---|
| 170 | """ |
---|
| 171 | Return parameter range string. |
---|
| 172 | |
---|
| 173 | E.g., " Gold .....|.... 5.2043 in [2,7]" |
---|
| 174 | """ |
---|
| 175 | bar = ['.']*10 |
---|
| 176 | lo,hi = self.range |
---|
| 177 | if numpy.isfinite(lo)and numpy.isfinite(hi): |
---|
| 178 | portion = (self.value-lo)/(hi-lo) |
---|
| 179 | if portion < 0: portion = 0. |
---|
| 180 | elif portion >= 1: portion = 0.99999999 |
---|
| 181 | barpos = int(math.floor(portion*len(bar))) |
---|
| 182 | bar[barpos] = '|' |
---|
| 183 | bar = "".join(bar) |
---|
| 184 | lostr = "[%g"%lo if numpy.isfinite(lo) else "(-inf" |
---|
| 185 | histr = "%g]"%hi if numpy.isfinite(hi) else "inf)" |
---|
| 186 | valstr = format_uncertainty(self.value, self.stderr) |
---|
| 187 | model_name = str(None) |
---|
| 188 | if self.model is not None: |
---|
| 189 | model_name = self.model.name |
---|
| 190 | data_name = str(None) |
---|
| 191 | if self.data is not None: |
---|
| 192 | data_name = self.data.name |
---|
| 193 | |
---|
| 194 | return "%25s %s %s in %s,%s, %s, %s" % (self.name,bar,valstr,lostr,histr, |
---|
| 195 | model_name, data_name) |
---|
| 196 | def __repr__(self): |
---|
| 197 | #return "FitParameter('%s')"%self.name |
---|
| 198 | return str(self.__class__) |
---|
| 199 | |
---|
[dfd3577] | 200 | class MyAssembly(Assembly): |
---|
| 201 | def __init__(self, models, curr_thread=None): |
---|
| 202 | Assembly.__init__(self, models) |
---|
| 203 | self.curr_thread = curr_thread |
---|
[bd5ac39] | 204 | self.chisq = None |
---|
| 205 | self._cancel = False |
---|
[e03a14e] | 206 | self.theory = None |
---|
[dfd3577] | 207 | |
---|
[1cff677] | 208 | def fit_parameters(self): |
---|
| 209 | """ |
---|
| 210 | Return an alphabetical list of the fitting parameters. |
---|
| 211 | |
---|
| 212 | This function is called once at the beginning of a fit, |
---|
| 213 | and serves as a convenient place to precalculate what |
---|
| 214 | can be precalculated such as the set of fitting parameters |
---|
| 215 | and the parameter expressions evaluator. |
---|
| 216 | """ |
---|
| 217 | self.parameterset.setprefix() |
---|
| 218 | self._fitparameters = self.parameterset.fitted |
---|
| 219 | self._restraints = self.parameterset.restrained |
---|
| 220 | pars = self.parameterset.flatten() |
---|
| 221 | context = self.parameterset.gather_context() |
---|
| 222 | self._fitexpression = park.expression.build_eval(pars,context) |
---|
| 223 | #print "constraints",self._fitexpression.__doc__ |
---|
| 224 | |
---|
[cef2da8] | 225 | self._fitparameters.sort(lambda a,b: cmp(a.path,b.path)) |
---|
[1cff677] | 226 | # Convert to fitparameter a object |
---|
| 227 | |
---|
| 228 | fitpars = [SansFitParameter(p.path,p.range,p.value, p.model, p.data) |
---|
| 229 | for p in self._fitparameters] |
---|
| 230 | #print "fitpars", fitpars |
---|
| 231 | return fitpars |
---|
| 232 | |
---|
| 233 | def all_results(self, result): |
---|
| 234 | """ |
---|
| 235 | Extend result from the fit with the calculated parameters. |
---|
| 236 | """ |
---|
| 237 | calcpars = [SansFitParameter(p.path,p.range,p.value, p.model, p.data) |
---|
| 238 | for p in self.parameterset.computed] |
---|
| 239 | result.parameters += calcpars |
---|
[e03a14e] | 240 | result.theory = self.theory |
---|
[1cff677] | 241 | |
---|
[dfd3577] | 242 | def eval(self): |
---|
| 243 | """ |
---|
| 244 | Recalculate the theory functions, and from them, the |
---|
| 245 | residuals and chisq. |
---|
| 246 | |
---|
| 247 | :note: Call this after the parameters have been updated. |
---|
| 248 | """ |
---|
| 249 | # Handle abort from a separate thread. |
---|
| 250 | self._cancel = False |
---|
| 251 | if self.curr_thread != None: |
---|
| 252 | try: |
---|
| 253 | self.curr_thread.isquit() |
---|
| 254 | except: |
---|
| 255 | self._cancel = True |
---|
| 256 | |
---|
| 257 | # Evaluate the computed parameters |
---|
[bd5ac39] | 258 | try: |
---|
| 259 | self._fitexpression() |
---|
| 260 | except NameError: |
---|
| 261 | pass |
---|
[dfd3577] | 262 | |
---|
| 263 | # Check that the resulting parameters are in a feasible region. |
---|
| 264 | if not self.isfeasible(): return numpy.inf |
---|
| 265 | |
---|
| 266 | resid = [] |
---|
| 267 | k = len(self._fitparameters) |
---|
| 268 | for m in self.parts: |
---|
| 269 | # In order to support abort, need to be able to propagate an |
---|
| 270 | # external abort signal from self.abort() into an abort signal |
---|
| 271 | # for the particular model. Can't see a way to do this which |
---|
| 272 | # doesn't involve setting a state variable. |
---|
| 273 | self._current_model = m |
---|
| 274 | if self._cancel: return numpy.inf |
---|
| 275 | if m.isfitted and m.weight != 0: |
---|
[e03a14e] | 276 | m.residuals, self.theory = m.fitness.residuals() |
---|
[dfd3577] | 277 | N = len(m.residuals) |
---|
| 278 | m.degrees_of_freedom = N-k if N>k else 1 |
---|
| 279 | m.chisq = numpy.sum(m.residuals**2) |
---|
| 280 | resid.append(m.weight*m.residuals) |
---|
| 281 | self.residuals = numpy.hstack(resid) |
---|
| 282 | N = len(self.residuals) |
---|
| 283 | self.degrees_of_freedom = N-k if N>k else 1 |
---|
| 284 | self.chisq = numpy.sum(self.residuals**2) |
---|
| 285 | return self.chisq |
---|
| 286 | |
---|
[4c718654] | 287 | class ParkFit(FitEngine): |
---|
[7705306] | 288 | """ |
---|
[aa36f96] | 289 | ParkFit performs the Fit.This class can be used as follow: |
---|
| 290 | #Do the fit Park |
---|
| 291 | create an engine: engine = ParkFit() |
---|
| 292 | Use data must be of type plottable |
---|
| 293 | Use a sans model |
---|
| 294 | |
---|
| 295 | Add data with a dictionnary of FitArrangeList where Uid is a key and data |
---|
| 296 | is saved in FitArrange object. |
---|
| 297 | engine.set_data(data,Uid) |
---|
| 298 | |
---|
| 299 | Set model parameter "M1"= model.name add {model.parameter.name:value}. |
---|
| 300 | |
---|
| 301 | :note: Set_param() if used must always preceded set_model() |
---|
| 302 | for the fit to be performed. |
---|
| 303 | engine.set_param( model,"M1", {'A':2,'B':4}) |
---|
| 304 | |
---|
[b2f25dc5] | 305 | Add model with a dictionnary of FitArrangeList{} where Uid is a key |
---|
| 306 | and model |
---|
[aa36f96] | 307 | is save in FitArrange object. |
---|
| 308 | engine.set_model(model,Uid) |
---|
| 309 | |
---|
| 310 | engine.fit return chisqr,[model.parameter 1,2,..],[[err1....][..err2...]] |
---|
| 311 | chisqr1, out1, cov1=engine.fit({model.parameter.name:value},qmin,qmax) |
---|
| 312 | |
---|
| 313 | :note: {model.parameter.name:value} is ignored in fit function since |
---|
[792db7d5] | 314 | the user should make sure to call set_param himself. |
---|
[aa36f96] | 315 | |
---|
[7705306] | 316 | """ |
---|
[916a15f] | 317 | def __init__(self): |
---|
[792db7d5] | 318 | """ |
---|
[aa36f96] | 319 | Creates a dictionary (self.fitArrangeList={})of FitArrange elements |
---|
| 320 | with Uid as keys |
---|
[792db7d5] | 321 | """ |
---|
[b2f25dc5] | 322 | FitEngine.__init__(self) |
---|
| 323 | self.fit_arrange_dict = {} |
---|
| 324 | self.param_list = [] |
---|
[37d9521] | 325 | |
---|
[dfd3577] | 326 | def create_assembly(self, curr_thread): |
---|
[7705306] | 327 | """ |
---|
[b2f25dc5] | 328 | Extract sansmodel and sansdata from |
---|
| 329 | self.FitArrangelist ={Uid:FitArrange} |
---|
| 330 | Create parkmodel and park data ,form a list couple of parkmodel |
---|
| 331 | and parkdata |
---|
[792db7d5] | 332 | create an assembly self.problem= park.Assembly([(parkmodel,parkdata)]) |
---|
[7705306] | 333 | """ |
---|
[89f3b66] | 334 | mylist = [] |
---|
[b2f25dc5] | 335 | #listmodel = [] |
---|
| 336 | #i = 0 |
---|
[89f3b66] | 337 | fitproblems = [] |
---|
[b2f25dc5] | 338 | for fproblem in self.fit_arrange_dict.itervalues(): |
---|
[89f3b66] | 339 | if fproblem.get_to_fit() == 1: |
---|
[a9e04aa] | 340 | fitproblems.append(fproblem) |
---|
[89f3b66] | 341 | if len(fitproblems) == 0: |
---|
[a9e04aa] | 342 | raise RuntimeError, "No Assembly scheduled for Park fitting." |
---|
| 343 | return |
---|
| 344 | for item in fitproblems: |
---|
| 345 | parkmodel = item.get_model() |
---|
[9e85792] | 346 | for p in parkmodel.parameterset: |
---|
[fe886ee] | 347 | ## does not allow status change for constraint parameters |
---|
[89f3b66] | 348 | if p.status != 'computed': |
---|
[b2f25dc5] | 349 | if p.get_name()in item.pars: |
---|
| 350 | ## make parameters selected for |
---|
| 351 | #fit will be between boundaries |
---|
[89f3b66] | 352 | p.set(p.range) |
---|
[aed7c57] | 353 | else: |
---|
[89f3b66] | 354 | p.status = 'fixed' |
---|
[b2f25dc5] | 355 | data_list = item.get_data() |
---|
| 356 | parkdata = data_list |
---|
[89f3b66] | 357 | fitness = (parkmodel, parkdata) |
---|
[ca6d914] | 358 | mylist.append(fitness) |
---|
[dfd3577] | 359 | self.problem = MyAssembly(models=mylist, curr_thread=curr_thread) |
---|
[126a761] | 360 | |
---|
[dfd3577] | 361 | |
---|
[93de635d] | 362 | def fit(self, q=None, handler=None, curr_thread=None, ftol=1.49012e-8): |
---|
[7705306] | 363 | """ |
---|
[aa36f96] | 364 | Performs fit with park.fit module.It can perform fit with one model |
---|
| 365 | and a set of data, more than two fit of one model and sets of data or |
---|
[b2f25dc5] | 366 | fit with more than two model associated with their set of data and |
---|
| 367 | constraints |
---|
[aa36f96] | 368 | |
---|
[b2f25dc5] | 369 | :param pars: Dictionary of parameter names for the model and their |
---|
| 370 | values. |
---|
[aa36f96] | 371 | :param qmin: The minimum value of data's range to be fit |
---|
| 372 | :param qmax: The maximum value of data's range to be fit |
---|
| 373 | |
---|
[b2f25dc5] | 374 | :note: all parameter are ignored most of the time.Are just there |
---|
| 375 | to keep ScipyFit and ParkFit interface the same. |
---|
[aa36f96] | 376 | |
---|
| 377 | :return: result.fitness Value of the goodness of fit metric |
---|
[b2f25dc5] | 378 | :return: result.pvec list of parameter with the best value |
---|
| 379 | found during fitting |
---|
[aa36f96] | 380 | :return: result.cov Covariance matrix |
---|
| 381 | |
---|
[7705306] | 382 | """ |
---|
[dfd3577] | 383 | self.create_assembly(curr_thread=curr_thread) |
---|
[1cff677] | 384 | localfit = SansFitSimplex() |
---|
[93de635d] | 385 | localfit.ftol = ftol |
---|
[681f0dc] | 386 | |
---|
[916a15f] | 387 | # See `park.fitresult.FitHandler` for details. |
---|
[1cff677] | 388 | fitter = SansFitMC(localfit=localfit, start_points=1) |
---|
[681f0dc] | 389 | if handler == None: |
---|
[89f3b66] | 390 | handler = fitresult.ConsoleUpdate(improvement_delta=0.1) |
---|
[dfd3577] | 391 | result = fit.fit(self.problem, fitter=fitter, handler=handler) |
---|
[8296ff5] | 392 | self.problem.all_results(result) |
---|
[1cff677] | 393 | |
---|
[634ca14] | 394 | #print "park------", result.inputs |
---|
| 395 | #for (model, data) in result.inputs: |
---|
| 396 | # print model.name, data.name |
---|
| 397 | #for p in result.parameters: |
---|
| 398 | # print "simul ----", p , p.__class__, p.model.name, p.data.name |
---|
[1cff677] | 399 | |
---|
[89f3b66] | 400 | if result != None: |
---|
| 401 | if q != None: |
---|
[fd6b789] | 402 | q.put(result) |
---|
| 403 | return q |
---|
[48882d1] | 404 | return result |
---|
[ee5b04c] | 405 | else: |
---|
| 406 | raise ValueError, "SVD did not converge" |
---|
[aa36f96] | 407 | |
---|