source: trunk/python/asapfitter.py@ 2014

Last change on this file since 2014 was 1938, checked in by Malte Marquarding, 14 years ago

merge from branches/asap4casa3.1.0. Should be done the other way

  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 23.0 KB
Line 
1import _asap
2from asap.parameters import rcParams
3from asap.logging import asaplog, asaplog_post_dec
4from asap.utils import _n_bools, mask_and
5
6
7class fitter:
8 """
9 The fitting class for ASAP.
10 """
11 def __init__(self):
12 """
13 Create a fitter object. No state is set.
14 """
15 self.fitter = _asap.fitter()
16 self.x = None
17 self.y = None
18 self.mask = None
19 self.fitfunc = None
20 self.fitfuncs = None
21 self.fitted = False
22 self.data = None
23 self.components = 0
24 self._fittedrow = 0
25 self._p = None
26 self._selection = None
27 self.uselinear = False
28
29 def set_data(self, xdat, ydat, mask=None):
30 """
31 Set the absissa and ordinate for the fit. Also set the mask
32 indicationg valid points.
33 This can be used for data vectors retrieved from a scantable.
34 For scantable fitting use 'fitter.set_scan(scan, mask)'.
35 Parameters:
36 xdat: the abcissa values
37 ydat: the ordinate values
38 mask: an optional mask
39
40 """
41 self.fitted = False
42 self.x = xdat
43 self.y = ydat
44 if mask == None:
45 self.mask = _n_bools(len(xdat), True)
46 else:
47 self.mask = mask
48 return
49
50 @asaplog_post_dec
51 def set_scan(self, thescan=None, mask=None):
52 """
53 Set the 'data' (a scantable) of the fitter.
54 Parameters:
55 thescan: a scantable
56 mask: a msk retrieved from the scantable
57 """
58 if not thescan:
59 msg = "Please give a correct scan"
60 raise TypeError(msg)
61 self.fitted = False
62 self.data = thescan
63 self.mask = None
64 if mask is None:
65 self.mask = _n_bools(self.data.nchan(), True)
66 else:
67 self.mask = mask
68 return
69
70 @asaplog_post_dec
71 def set_function(self, **kwargs):
72 """
73 Set the function to be fit.
74 Parameters:
75 poly: use a polynomial of the order given with nonlinear least squares fit
76 lpoly: use polynomial of the order given with linear least squares fit
77 gauss: fit the number of gaussian specified
78 lorentz: fit the number of lorentzian specified
79 Example:
80 fitter.set_function(poly=3) # will fit a 3rd order polynomial via nonlinear method
81 fitter.set_function(lpoly=3) # will fit a 3rd order polynomial via linear method
82 fitter.set_function(gauss=2) # will fit two gaussians
83 fitter.set_function(lorentz=2) # will fit two lorentzians
84 """
85 #default poly order 0
86 n=0
87 if kwargs.has_key('poly'):
88 self.fitfunc = 'poly'
89 self.fitfuncs = ['poly']
90 n = kwargs.get('poly')
91 self.components = [n+1]
92 self.uselinear = False
93 elif kwargs.has_key('lpoly'):
94 self.fitfunc = 'poly'
95 self.fitfuncs = ['lpoly']
96 n = kwargs.get('lpoly')
97 self.components = [n+1]
98 self.uselinear = True
99 elif kwargs.has_key('gauss'):
100 n = kwargs.get('gauss')
101 self.fitfunc = 'gauss'
102 self.fitfuncs = [ 'gauss' for i in range(n) ]
103 self.components = [ 3 for i in range(n) ]
104 self.uselinear = False
105 elif kwargs.has_key('lorentz'):
106 n = kwargs.get('lorentz')
107 self.fitfunc = 'lorentz'
108 self.fitfuncs = [ 'lorentz' for i in range(n) ]
109 self.components = [ 3 for i in range(n) ]
110 self.uselinear = False
111 else:
112 msg = "Invalid function type."
113 raise TypeError(msg)
114
115 self.fitter.setexpression(self.fitfunc,n)
116 self.fitted = False
117 return
118
119 @asaplog_post_dec
120 def fit(self, row=0, estimate=False):
121 """
122 Execute the actual fitting process. All the state has to be set.
123 Parameters:
124 row: specify the row in the scantable
125 estimate: auto-compute an initial parameter set (default False)
126 This can be used to compute estimates even if fit was
127 called before.
128 Example:
129 s = scantable('myscan.asap')
130 s.set_cursor(thepol=1) # select second pol
131 f = fitter()
132 f.set_scan(s)
133 f.set_function(poly=0)
134 f.fit(row=0) # fit first row
135 """
136 if ((self.x is None or self.y is None) and self.data is None) \
137 or self.fitfunc is None:
138 msg = "Fitter not yet initialised. Please set data & fit function"
139 raise RuntimeError(msg)
140
141 else:
142 if self.data is not None:
143 self.x = self.data._getabcissa(row)
144 self.y = self.data._getspectrum(row)
145 self.mask = mask_and(self.mask, self.data._getmask(row))
146 asaplog.push("Fitting:")
147 i = row
148 out = "Scan[%d] Beam[%d] IF[%d] Pol[%d] Cycle[%d]" % (self.data.getscan(i),
149 self.data.getbeam(i),
150 self.data.getif(i),
151 self.data.getpol(i),
152 self.data.getcycle(i))
153 asaplog.push(out,False)
154 self.fitter.setdata(self.x, self.y, self.mask)
155 if self.fitfunc == 'gauss' or self.fitfunc == 'lorentz':
156 ps = self.fitter.getparameters()
157 if len(ps) == 0 or estimate:
158 self.fitter.estimate()
159 fxdpar = list(self.fitter.getfixedparameters())
160 if len(fxdpar) and fxdpar.count(0) == 0:
161 raise RuntimeError,"No point fitting, if all parameters are fixed."
162 if self.uselinear:
163 converged = self.fitter.lfit()
164 else:
165 converged = self.fitter.fit()
166 if not converged:
167 raise RuntimeError,"Fit didn't converge."
168 self._fittedrow = row
169 self.fitted = True
170 return
171
172 def store_fit(self, filename=None):
173 """
174 Save the fit parameters.
175 Parameters:
176 filename: if specified save as an ASCII file, if None (default)
177 store it in the scnatable
178 """
179 if self.fitted and self.data is not None:
180 pars = list(self.fitter.getparameters())
181 fixed = list(self.fitter.getfixedparameters())
182 from asap.asapfit import asapfit
183 fit = asapfit()
184 fit.setparameters(pars)
185 fit.setfixedparameters(fixed)
186 fit.setfunctions(self.fitfuncs)
187 fit.setcomponents(self.components)
188 fit.setframeinfo(self.data._getcoordinfo())
189 if filename is not None:
190 import os
191 filename = os.path.expandvars(os.path.expanduser(filename))
192 if os.path.exists(filename):
193 raise IOError("File '%s' exists." % filename)
194 fit.save(filename)
195 else:
196 self.data._addfit(fit,self._fittedrow)
197
198 @asaplog_post_dec
199 def set_parameters(self,*args,**kwargs):
200 """
201 Set the parameters to be fitted.
202 Parameters:
203 params: a vector of parameters
204 fixed: a vector of which parameters are to be held fixed
205 (default is none)
206 component: in case of multiple gaussians/lorentzians,
207 the index of the component
208 """
209 component = None
210 fixed = None
211 params = None
212
213 if len(args) and isinstance(args[0],dict):
214 kwargs = args[0]
215 if kwargs.has_key("fixed"): fixed = kwargs["fixed"]
216 if kwargs.has_key("params"): params = kwargs["params"]
217 if len(args) == 2 and isinstance(args[1], int):
218 component = args[1]
219 if self.fitfunc is None:
220 msg = "Please specify a fitting function first."
221 raise RuntimeError(msg)
222 if (self.fitfunc == "gauss" or self.fitfunc == 'lorentz') and component is not None:
223 if not self.fitted and sum(self.fitter.getparameters()) == 0:
224 pars = _n_bools(len(self.components)*3, False)
225 fxd = _n_bools(len(pars), False)
226 else:
227 pars = list(self.fitter.getparameters())
228 fxd = list(self.fitter.getfixedparameters())
229 i = 3*component
230 pars[i:i+3] = params
231 fxd[i:i+3] = fixed
232 params = pars
233 fixed = fxd
234 self.fitter.setparameters(params)
235 if fixed is not None:
236 self.fitter.setfixedparameters(fixed)
237 return
238
239 @asaplog_post_dec
240 def set_gauss_parameters(self, peak, centre, fwhm,
241 peakfixed=0, centrefixed=0,
242 fwhmfixed=0,
243 component=0):
244 """
245 Set the Parameters of a 'Gaussian' component, set with set_function.
246 Parameters:
247 peak, centre, fwhm: The gaussian parameters
248 peakfixed,
249 centrefixed,
250 fwhmfixed: Optional parameters to indicate if
251 the paramters should be held fixed during
252 the fitting process. The default is to keep
253 all parameters flexible.
254 component: The number of the component (Default is the
255 component 0)
256 """
257 if self.fitfunc != "gauss":
258 msg = "Function only operates on Gaussian components."
259 raise ValueError(msg)
260 if 0 <= component < len(self.components):
261 d = {'params':[peak, centre, fwhm],
262 'fixed':[peakfixed, centrefixed, fwhmfixed]}
263 self.set_parameters(d, component)
264 else:
265 msg = "Please select a valid component."
266 raise ValueError(msg)
267
268 @asaplog_post_dec
269 def set_lorentz_parameters(self, peak, centre, fwhm,
270 peakfixed=0, centrefixed=0,
271 fwhmfixed=0,
272 component=0):
273 """
274 Set the Parameters of a 'Lorentzian' component, set with set_function.
275 Parameters:
276 peak, centre, fwhm: The lorentzian parameters
277 peakfixed,
278 centrefixed,
279 fwhmfixed: Optional parameters to indicate if
280 the paramters should be held fixed during
281 the fitting process. The default is to keep
282 all parameters flexible.
283 component: The number of the component (Default is the
284 component 0)
285 """
286 if self.fitfunc != "lorentz":
287 msg = "Function only operates on Lorentzian components."
288 raise ValueError(msg)
289 if 0 <= component < len(self.components):
290 d = {'params':[peak, centre, fwhm],
291 'fixed':[peakfixed, centrefixed, fwhmfixed]}
292 self.set_parameters(d, component)
293 else:
294 msg = "Please select a valid component."
295 raise ValueError(msg)
296
297 def get_area(self, component=None):
298 """
299 Return the area under the fitted gaussian/lorentzian component.
300 Parameters:
301 component: the gaussian/lorentzian component selection,
302 default (None) is the sum of all components
303 Note:
304 This will only work for gaussian/lorentzian fits.
305 """
306 if not self.fitted: return
307 if self.fitfunc == "gauss" or self.fitfunc == "lorentz":
308 pars = list(self.fitter.getparameters())
309 from math import log,pi,sqrt
310 if self.fitfunc == "gauss":
311 fac = sqrt(pi/log(16.0))
312 elif self.fitfunc == "lorentz":
313 fac = pi/2.0
314 areas = []
315 for i in range(len(self.components)):
316 j = i*3
317 cpars = pars[j:j+3]
318 areas.append(fac * cpars[0] * cpars[2])
319 else:
320 return None
321 if component is not None:
322 return areas[component]
323 else:
324 return sum(areas)
325
326 @asaplog_post_dec
327 def get_errors(self, component=None):
328 """
329 Return the errors in the parameters.
330 Parameters:
331 component: get the errors for the specified component
332 only, default is all components
333 """
334 if not self.fitted:
335 msg = "Not yet fitted."
336 raise RuntimeError(msg)
337 errs = list(self.fitter.geterrors())
338 cerrs = errs
339 if component is not None:
340 if self.fitfunc == "gauss" or self.fitfunc == "lorentz":
341 i = 3*component
342 if i < len(errs):
343 cerrs = errs[i:i+3]
344 return cerrs
345
346
347 @asaplog_post_dec
348 def get_parameters(self, component=None, errors=False):
349 """
350 Return the fit paramters.
351 Parameters:
352 component: get the parameters for the specified component
353 only, default is all components
354 """
355 if not self.fitted:
356 msg = "Not yet fitted."
357 raise RuntimeError(msg)
358 pars = list(self.fitter.getparameters())
359 fixed = list(self.fitter.getfixedparameters())
360 errs = list(self.fitter.geterrors())
361 area = []
362 if component is not None:
363 if self.fitfunc == "gauss" or self.fitfunc == "lorentz":
364 i = 3*component
365 cpars = pars[i:i+3]
366 cfixed = fixed[i:i+3]
367 cerrs = errs[i:i+3]
368 a = self.get_area(component)
369 area = [a for i in range(3)]
370 else:
371 cpars = pars
372 cfixed = fixed
373 cerrs = errs
374 else:
375 cpars = pars
376 cfixed = fixed
377 cerrs = errs
378 if self.fitfunc == "gauss" or self.fitfunc == "lorentz":
379 for c in range(len(self.components)):
380 a = self.get_area(c)
381 area += [a for i in range(3)]
382 fpars = self._format_pars(cpars, cfixed, errors and cerrs, area)
383 asaplog.push(fpars)
384 return {'params':cpars, 'fixed':cfixed, 'formatted': fpars,
385 'errors':cerrs}
386
387 def _format_pars(self, pars, fixed, errors, area):
388 out = ''
389 if self.fitfunc == 'poly':
390 c = 0
391 for i in range(len(pars)):
392 fix = ""
393 if len(fixed) and fixed[i]: fix = "(fixed)"
394 if errors :
395 out += ' p%d%s= %3.6f (%1.6f),' % (c,fix,pars[i], errors[i])
396 else:
397 out += ' p%d%s= %3.6f,' % (c,fix,pars[i])
398 c+=1
399 out = out[:-1] # remove trailing ','
400 elif self.fitfunc == 'gauss' or self.fitfunc == 'lorentz':
401 i = 0
402 c = 0
403 aunit = ''
404 ounit = ''
405 if self.data:
406 aunit = self.data.get_unit()
407 ounit = self.data.get_fluxunit()
408 while i < len(pars):
409 if len(area):
410 out += ' %2d: peak = %3.3f %s , centre = %3.3f %s, FWHM = %3.3f %s\n area = %3.3f %s %s\n' % (c,pars[i],ounit,pars[i+1],aunit,pars[i+2],aunit, area[i],ounit,aunit)
411 else:
412 out += ' %2d: peak = %3.3f %s , centre = %3.3f %s, FWHM = %3.3f %s\n' % (c,pars[i],ounit,pars[i+1],aunit,pars[i+2],aunit,ounit,aunit)
413 c+=1
414 i+=3
415 return out
416
417
418 @asaplog_post_dec
419 def get_estimate(self):
420 """
421 Return the parameter estimates (for non-linear functions).
422 """
423 pars = self.fitter.getestimate()
424 fixed = self.fitter.getfixedparameters()
425 asaplog.push(self._format_pars(pars,fixed,None,None))
426 return pars
427
428 @asaplog_post_dec
429 def get_residual(self):
430 """
431 Return the residual of the fit.
432 """
433 if not self.fitted:
434 msg = "Not yet fitted."
435 raise RuntimeError(msg)
436 return self.fitter.getresidual()
437
438 @asaplog_post_dec
439 def get_chi2(self):
440 """
441 Return chi^2.
442 """
443 if not self.fitted:
444 msg = "Not yet fitted."
445 raise RuntimeError(msg)
446 ch2 = self.fitter.getchi2()
447 asaplog.push( 'Chi^2 = %3.3f' % (ch2) )
448 return ch2
449
450 @asaplog_post_dec
451 def get_fit(self):
452 """
453 Return the fitted ordinate values.
454 """
455 if not self.fitted:
456 msg = "Not yet fitted."
457 raise RuntimeError(msg)
458 return self.fitter.getfit()
459
460 @asaplog_post_dec
461 def commit(self):
462 """
463 Return a new scan where the fits have been commited (subtracted)
464 """
465 if not self.fitted:
466 msg = "Not yet fitted."
467 raise RuntimeError(msg)
468 from asap import scantable
469 if not isinstance(self.data, scantable):
470 msg = "Not a scantable"
471 raise TypeError(msg)
472 scan = self.data.copy()
473 scan._setspectrum(self.fitter.getresidual())
474 return scan
475
476 @asaplog_post_dec
477 def plot(self, residual=False, components=None, plotparms=False,
478 filename=None):
479 """
480 Plot the last fit.
481 Parameters:
482 residual: an optional parameter indicating if the residual
483 should be plotted (default 'False')
484 components: a list of components to plot, e.g [0,1],
485 -1 plots the total fit. Default is to only
486 plot the total fit.
487 plotparms: Inidicates if the parameter values should be present
488 on the plot
489 """
490 if not self.fitted:
491 return
492 if not self._p or self._p.is_dead:
493 if rcParams['plotter.gui']:
494 from asap.asaplotgui import asaplotgui as asaplot
495 else:
496 from asap.asaplot import asaplot
497 self._p = asaplot()
498 self._p.hold()
499 self._p.clear()
500 self._p.set_panels()
501 self._p.palette(0)
502 tlab = 'Spectrum'
503 xlab = 'Abcissa'
504 ylab = 'Ordinate'
505 from numpy import ma,logical_not,logical_and,array
506 m = self.mask
507 if self.data:
508 tlab = self.data._getsourcename(self._fittedrow)
509 xlab = self.data._getabcissalabel(self._fittedrow)
510 m = logical_and(self.mask,
511 array(self.data._getmask(self._fittedrow),
512 copy=False))
513
514 ylab = self.data._get_ordinate_label()
515
516 colours = ["#777777","#dddddd","red","orange","purple","green","magenta", "cyan"]
517 nomask=True
518 for i in range(len(m)):
519 nomask = nomask and m[i]
520 label0='Masked Region'
521 label1='Spectrum'
522 if ( nomask ):
523 label0=label1
524 else:
525 y = ma.masked_array( self.y, mask = m )
526 self._p.palette(1,colours)
527 self._p.set_line( label = label1 )
528 self._p.plot( self.x, y )
529 self._p.palette(0,colours)
530 self._p.set_line(label=label0)
531 y = ma.masked_array(self.y,mask=logical_not(m))
532 self._p.plot(self.x, y)
533 if residual:
534 self._p.palette(7)
535 self._p.set_line(label='Residual')
536 y = ma.masked_array(self.get_residual(),
537 mask=logical_not(m))
538 self._p.plot(self.x, y)
539 self._p.palette(2)
540 if components is not None:
541 cs = components
542 if isinstance(components,int): cs = [components]
543 if plotparms:
544 self._p.text(0.15,0.15,str(self.get_parameters()['formatted']),size=8)
545 n = len(self.components)
546 self._p.palette(3)
547 for c in cs:
548 if 0 <= c < n:
549 lab = self.fitfuncs[c]+str(c)
550 self._p.set_line(label=lab)
551 y = ma.masked_array(self.fitter.evaluate(c),
552 mask=logical_not(m))
553
554 self._p.plot(self.x, y)
555 elif c == -1:
556 self._p.palette(2)
557 self._p.set_line(label="Total Fit")
558 y = ma.masked_array(self.fitter.getfit(),
559 mask=logical_not(m))
560 self._p.plot(self.x, y)
561 else:
562 self._p.palette(2)
563 self._p.set_line(label='Fit')
564 y = ma.masked_array(self.fitter.getfit(),
565 mask=logical_not(m))
566 self._p.plot(self.x, y)
567 xlim=[min(self.x),max(self.x)]
568 self._p.axes.set_xlim(xlim)
569 self._p.set_axes('xlabel',xlab)
570 self._p.set_axes('ylabel',ylab)
571 self._p.set_axes('title',tlab)
572 self._p.release()
573 if (not rcParams['plotter.gui']):
574 self._p.save(filename)
575
576 @asaplog_post_dec
577 def auto_fit(self, insitu=None, plot=False):
578 """
579 Return a scan where the function is applied to all rows for
580 all Beams/IFs/Pols.
581
582 """
583 from asap import scantable
584 if not isinstance(self.data, scantable) :
585 msg = "Data is not a scantable"
586 raise TypeError(msg)
587 if insitu is None: insitu = rcParams['insitu']
588 if not insitu:
589 scan = self.data.copy()
590 else:
591 scan = self.data
592 rows = xrange(scan.nrow())
593 # Save parameters of baseline fits as a class attribute.
594 # NOTICE: This does not reflect changes in scantable!
595 if len(rows) > 0: self.blpars=[]
596 asaplog.push("Fitting:")
597 for r in rows:
598 out = " Scan[%d] Beam[%d] IF[%d] Pol[%d] Cycle[%d]" % (scan.getscan(r),
599 scan.getbeam(r),
600 scan.getif(r),
601 scan.getpol(r),
602 scan.getcycle(r))
603 asaplog.push(out, False)
604 self.x = scan._getabcissa(r)
605 self.y = scan._getspectrum(r)
606 self.mask = mask_and(self.mask, scan._getmask(r))
607 self.data = None
608 self.fit()
609 x = self.get_parameters()
610 fpar = self.get_parameters()
611 if plot:
612 self.plot(residual=True)
613 x = raw_input("Accept fit ([y]/n): ")
614 if x.upper() == 'N':
615 self.blpars.append(None)
616 continue
617 scan._setspectrum(self.fitter.getresidual(), r)
618 self.blpars.append(fpar)
619 if plot:
620 self._p.unmap()
621 self._p = None
622 return scan
Note: See TracBrowser for help on using the repository browser.