source: trunk/python/asapfitter.py@ 1393

Last change on this file since 1393 was 1391, checked in by Malte Marquarding, 17 years ago

merge from alma branch to get alma/GBT support. Commented out fluxUnit changes as they are using a chnaged interface to PKSreader/writer. Also commented out progress meter related code.

  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 21.2 KB
Line 
1import _asap
2from asap import rcParams
3from asap import print_log
4from asap import _n_bools
5
6class fitter:
7 """
8 The fitting class for ASAP.
9 """
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 def set_scan(self, thescan=None, mask=None):
51 """
52 Set the 'data' (a scantable) of the fitter.
53 Parameters:
54 thescan: a scantable
55 mask: a msk retireved from the scantable
56 """
57 if not thescan:
58 msg = "Please give a correct scan"
59 if rcParams['verbose']:
60 print msg
61 return
62 else:
63 raise TypeError(msg)
64 self.fitted = False
65 self.data = thescan
66 self.mask = None
67 if mask is None:
68 self.mask = _n_bools(self.data.nchan(), True)
69 else:
70 self.mask = mask
71 return
72
73 def set_function(self, **kwargs):
74 """
75 Set the function to be fit.
76 Parameters:
77 poly: use a polynomial of the order given with nonlinear least squares fit
78 lpoly: use polynomial of the order given with linear least squares fit
79 gauss: fit the number of gaussian specified
80 Example:
81 fitter.set_function(gauss=2) # will fit two gaussians
82 fitter.set_function(poly=3) # will fit a 3rd order polynomial via nonlinear method
83 fitter.set_function(lpoly=3) # will fit a 3rd order polynomial via linear method
84 """
85 #default poly order 0
86 n=0
87 if kwargs.has_key('poly'):
88 self.fitfunc = 'poly'
89 n = kwargs.get('poly')
90 self.components = [n]
91 self.uselinear = False
92 elif kwargs.has_key('lpoly'):
93 self.fitfunc = 'poly'
94 n = kwargs.get('lpoly')
95 self.components = [n]
96 self.uselinear = True
97 elif kwargs.has_key('gauss'):
98 n = kwargs.get('gauss')
99 self.fitfunc = 'gauss'
100 self.fitfuncs = [ 'gauss' for i in range(n) ]
101 self.components = [ 3 for i in range(n) ]
102 self.uselinear = False
103 else:
104 msg = "Invalid function type."
105 if rcParams['verbose']:
106 print msg
107 return
108 else:
109 raise TypeError(msg)
110
111 self.fitter.setexpression(self.fitfunc,n)
112 self.fitted = False
113 return
114
115 def fit(self, row=0, estimate=False):
116 """
117 Execute the actual fitting process. All the state has to be set.
118 Parameters:
119 row: specify the row in the scantable
120 estimate: auto-compute an initial parameter set (default False)
121 This can be used to compute estimates even if fit was
122 called before.
123 Example:
124 s = scantable('myscan.asap')
125 s.set_cursor(thepol=1) # select second pol
126 f = fitter()
127 f.set_scan(s)
128 f.set_function(poly=0)
129 f.fit(row=0) # fit first row
130 """
131 if ((self.x is None or self.y is None) and self.data is None) \
132 or self.fitfunc is None:
133 msg = "Fitter not yet initialised. Please set data & fit function"
134 if rcParams['verbose']:
135 print msg
136 return
137 else:
138 raise RuntimeError(msg)
139
140 else:
141 if self.data is not None:
142 self.x = self.data._getabcissa(row)
143 self.y = self.data._getspectrum(row)
144 from asap import asaplog
145 asaplog.push("Fitting:")
146 i = row
147 out = "Scan[%d] Beam[%d] IF[%d] Pol[%d] Cycle[%d]" % (self.data.getscan(i),self.data.getbeam(i),self.data.getif(i),self.data.getpol(i), self.data.getcycle(i))
148 asaplog.push(out,False)
149 self.fitter.setdata(self.x, self.y, self.mask)
150 if self.fitfunc == 'gauss':
151 ps = self.fitter.getparameters()
152 if len(ps) == 0 or estimate:
153 self.fitter.estimate()
154 try:
155 fxdpar = list(self.fitter.getfixedparameters())
156 if len(fxdpar) and fxdpar.count(0) == 0:
157 raise RuntimeError,"No point fitting, if all parameters are fixed."
158 if self.uselinear:
159 converged = self.fitter.lfit()
160 else:
161 converged = self.fitter.fit()
162 if not converged:
163 raise RuntimeError,"Fit didn't converge."
164 except RuntimeError, msg:
165 if rcParams['verbose']:
166 print msg
167 else:
168 raise
169 self._fittedrow = row
170 self.fitted = True
171 print_log()
172 return
173
174 def store_fit(self, filename=None):
175 """
176 Save the fit parameters.
177 Parameters:
178 filename: if specified save as an ASCII file, if None (default)
179 store it in the scnatable
180 """
181 if self.fitted and self.data is not None:
182 pars = list(self.fitter.getparameters())
183 fixed = list(self.fitter.getfixedparameters())
184 from asap.asapfit import asapfit
185 fit = asapfit()
186 fit.setparameters(pars)
187 fit.setfixedparameters(fixed)
188 fit.setfunctions(self.fitfuncs)
189 fit.setcomponents(self.components)
190 fit.setframeinfo(self.data._getcoordinfo())
191 if filename is not None:
192 import os
193 filename = os.path.expandvars(os.path.expanduser(filename))
194 if os.path.exists(filename):
195 raise IOError("File '%s' exists." % filename)
196 fit.save(filename)
197 else:
198 self.data._addfit(fit,self._fittedrow)
199
200 #def set_parameters(self, params, fixed=None, component=None):
201 def set_parameters(self,*args,**kwargs):
202 """
203 Set the parameters to be fitted.
204 Parameters:
205 params: a vector of parameters
206 fixed: a vector of which parameters are to be held fixed
207 (default is none)
208 component: in case of multiple gaussians, the index of the
209 component
210 """
211 component = None
212 fixed = None
213 params = None
214
215 if len(args) and isinstance(args[0],dict):
216 kwargs = args[0]
217 if kwargs.has_key("fixed"): fixed = kwargs["fixed"]
218 if kwargs.has_key("params"): params = kwargs["params"]
219 if len(args) == 2 and isinstance(args[1], int):
220 component = args[1]
221 if self.fitfunc is None:
222 msg = "Please specify a fitting function first."
223 if rcParams['verbose']:
224 print msg
225 return
226 else:
227 raise RuntimeError(msg)
228 if self.fitfunc == "gauss" and component is not None:
229 if not self.fitted and sum(self.fitter.getparameters()) == 0:
230 pars = _n_bools(len(self.components)*3, False)
231 fxd = _n_bools(len(pars), False)
232 else:
233 pars = list(self.fitter.getparameters())
234 fxd = list(self.fitter.getfixedparameters())
235 i = 3*component
236 pars[i:i+3] = params
237 fxd[i:i+3] = fixed
238 params = pars
239 fixed = fxd
240 self.fitter.setparameters(params)
241 if fixed is not None:
242 self.fitter.setfixedparameters(fixed)
243 print_log()
244 return
245
246 def set_gauss_parameters(self, peak, centre, fwhm,
247 peakfixed=0, centerfixed=0,
248 fwhmfixed=0,
249 component=0):
250 """
251 Set the Parameters of a 'Gaussian' component, set with set_function.
252 Parameters:
253 peak, centre, fwhm: The gaussian parameters
254 peakfixed,
255 centerfixed,
256 fwhmfixed: Optional parameters to indicate if
257 the paramters should be held fixed during
258 the fitting process. The default is to keep
259 all parameters flexible.
260 component: The number of the component (Default is the
261 component 0)
262 """
263 if self.fitfunc != "gauss":
264 msg = "Function only operates on Gaussian components."
265 if rcParams['verbose']:
266 print msg
267 return
268 else:
269 raise ValueError(msg)
270 if 0 <= component < len(self.components):
271 d = {'params':[peak, centre, fwhm],
272 'fixed':[peakfixed, centerfixed, fwhmfixed]}
273 self.set_parameters(d, component)
274 else:
275 msg = "Please select a valid component."
276 if rcParams['verbose']:
277 print msg
278 return
279 else:
280 raise ValueError(msg)
281
282 def get_area(self, component=None):
283 """
284 Return the area under the fitted gaussian component.
285 Parameters:
286 component: the gaussian component selection,
287 default (None) is the sum of all components
288 Note:
289 This will only work for gaussian fits.
290 """
291 if not self.fitted: return
292 if self.fitfunc == "gauss":
293 pars = list(self.fitter.getparameters())
294 from math import log,pi,sqrt
295 fac = sqrt(pi/log(16.0))
296 areas = []
297 for i in range(len(self.components)):
298 j = i*3
299 cpars = pars[j:j+3]
300 areas.append(fac * cpars[0] * cpars[2])
301 else:
302 return None
303 if component is not None:
304 return areas[component]
305 else:
306 return sum(areas)
307
308 def get_errors(self, component=None):
309 """
310 Return the errors in the parameters.
311 Parameters:
312 component: get the errors for the specified component
313 only, default is all components
314 """
315 if not self.fitted:
316 msg = "Not yet fitted."
317 if rcParams['verbose']:
318 print msg
319 return
320 else:
321 raise RuntimeError(msg)
322 errs = list(self.fitter.geterrors())
323 cerrs = errs
324 if component is not None:
325 if self.fitfunc == "gauss":
326 i = 3*component
327 if i < len(errs):
328 cerrs = errs[i:i+3]
329 return cerrs
330
331 def get_parameters(self, component=None, errors=False):
332 """
333 Return the fit paramters.
334 Parameters:
335 component: get the parameters for the specified component
336 only, default is all components
337 """
338 if not self.fitted:
339 msg = "Not yet fitted."
340 if rcParams['verbose']:
341 print msg
342 return
343 else:
344 raise RuntimeError(msg)
345 pars = list(self.fitter.getparameters())
346 fixed = list(self.fitter.getfixedparameters())
347 errs = list(self.fitter.geterrors())
348 area = []
349 if component is not None:
350 if self.fitfunc == "gauss":
351 i = 3*component
352 cpars = pars[i:i+3]
353 cfixed = fixed[i:i+3]
354 cerrs = errs[i:i+3]
355 a = self.get_area(component)
356 area = [a for i in range(3)]
357 else:
358 cpars = pars
359 cfixed = fixed
360 cerrs = errs
361 else:
362 cpars = pars
363 cfixed = fixed
364 cerrs = errs
365 if self.fitfunc == "gauss":
366 for c in range(len(self.components)):
367 a = self.get_area(c)
368 area += [a for i in range(3)]
369 fpars = self._format_pars(cpars, cfixed, errors and cerrs, area)
370 if rcParams['verbose']:
371 print fpars
372 return {'params':cpars, 'fixed':cfixed, 'formatted': fpars,
373 'errors':cerrs}
374
375 def _format_pars(self, pars, fixed, errors, area):
376 out = ''
377 if self.fitfunc == 'poly':
378 c = 0
379 for i in range(len(pars)):
380 fix = ""
381 if len(fixed) and fixed[i]: fix = "(fixed)"
382 if errors :
383 out += ' p%d%s= %3.6f (%1.6f),' % (c,fix,pars[i], errors[i])
384 else:
385 out += ' p%d%s= %3.6f,' % (c,fix,pars[i])
386 c+=1
387 out = out[:-1] # remove trailing ','
388 elif self.fitfunc == 'gauss':
389 i = 0
390 c = 0
391 aunit = ''
392 ounit = ''
393 if self.data:
394 aunit = self.data.get_unit()
395 ounit = self.data.get_fluxunit()
396 while i < len(pars):
397 if len(area):
398 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)
399 else:
400 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)
401 c+=1
402 i+=3
403 return out
404
405 def get_estimate(self):
406 """
407 Return the parameter estimates (for non-linear functions).
408 """
409 pars = self.fitter.getestimate()
410 fixed = self.fitter.getfixedparameters()
411 if rcParams['verbose']:
412 print self._format_pars(pars,fixed,None)
413 return pars
414
415 def get_residual(self):
416 """
417 Return the residual of the fit.
418 """
419 if not self.fitted:
420 msg = "Not yet fitted."
421 if rcParams['verbose']:
422 print msg
423 return
424 else:
425 raise RuntimeError(msg)
426 return self.fitter.getresidual()
427
428 def get_chi2(self):
429 """
430 Return chi^2.
431 """
432 if not self.fitted:
433 msg = "Not yet fitted."
434 if rcParams['verbose']:
435 print msg
436 return
437 else:
438 raise RuntimeError(msg)
439 ch2 = self.fitter.getchi2()
440 if rcParams['verbose']:
441 print 'Chi^2 = %3.3f' % (ch2)
442 return ch2
443
444 def get_fit(self):
445 """
446 Return the fitted ordinate values.
447 """
448 if not self.fitted:
449 msg = "Not yet fitted."
450 if rcParams['verbose']:
451 print msg
452 return
453 else:
454 raise RuntimeError(msg)
455 return self.fitter.getfit()
456
457 def commit(self):
458 """
459 Return a new scan where the fits have been commited (subtracted)
460 """
461 if not self.fitted:
462 msg = "Not yet fitted."
463 if rcParams['verbose']:
464 print msg
465 return
466 else:
467 raise RuntimeError(msg)
468 from asap import scantable
469 if not isinstance(self.data, scantable):
470 msg = "Not a scantable"
471 if rcParams['verbose']:
472 print msg
473 return
474 else:
475 raise TypeError(msg)
476 scan = self.data.copy()
477 scan._setspectrum(self.fitter.getresidual())
478 print_log()
479 return scan
480
481 def plot(self, residual=False, components=None, plotparms=False, filename=None):
482 """
483 Plot the last fit.
484 Parameters:
485 residual: an optional parameter indicating if the residual
486 should be plotted (default 'False')
487 components: a list of components to plot, e.g [0,1],
488 -1 plots the total fit. Default is to only
489 plot the total fit.
490 plotparms: Inidicates if the parameter values should be present
491 on the plot
492 """
493 if not self.fitted:
494 return
495 if not self._p or self._p.is_dead:
496 if rcParams['plotter.gui']:
497 from asap.asaplotgui import asaplotgui as asaplot
498 else:
499 from asap.asaplot import asaplot
500 self._p = asaplot()
501 self._p.hold()
502 self._p.clear()
503 self._p.set_panels()
504 self._p.palette(0)
505 tlab = 'Spectrum'
506 xlab = 'Abcissa'
507 ylab = 'Ordinate'
508 from matplotlib.numerix import ma,logical_not,logical_and,array
509 m = self.mask
510 if self.data:
511 tlab = self.data._getsourcename(self._fittedrow)
512 xlab = self.data._getabcissalabel(self._fittedrow)
513 m = logical_and(self.mask,
514 array(self.data._getmask(self._fittedrow),
515 copy=False))
516
517 ylab = self.data._get_ordinate_label()
518
519 colours = ["#777777","#dddddd","red","orange","purple","green","magenta", "cyan"]
520 self._p.palette(0,colours)
521 self._p.set_line(label='Spectrum')
522 y = ma.masked_array(self.y,mask=logical_not(m))
523 self._p.plot(self.x, y)
524 if residual:
525 self._p.palette(1)
526 self._p.set_line(label='Residual')
527 y = ma.masked_array(self.get_residual(),
528 mask=logical_not(m))
529 self._p.plot(self.x, y)
530 self._p.palette(2)
531 if components is not None:
532 cs = components
533 if isinstance(components,int): cs = [components]
534 if plotparms:
535 self._p.text(0.15,0.15,str(self.get_parameters()['formatted']),size=8)
536 n = len(self.components)
537 self._p.palette(3)
538 for c in cs:
539 if 0 <= c < n:
540 lab = self.fitfuncs[c]+str(c)
541 self._p.set_line(label=lab)
542 y = ma.masked_array(self.fitter.evaluate(c),
543 mask=logical_not(m))
544
545 self._p.plot(self.x, y)
546 elif c == -1:
547 self._p.palette(2)
548 self._p.set_line(label="Total Fit")
549 y = ma.masked_array(self.fitter.getfit(),
550 mask=logical_not(m))
551 self._p.plot(self.x, y)
552 else:
553 self._p.palette(2)
554 self._p.set_line(label='Fit')
555 y = ma.masked_array(self.fitter.getfit(),
556 mask=logical_not(m))
557 self._p.plot(self.x, y)
558 xlim=[min(self.x),max(self.x)]
559 self._p.axes.set_xlim(xlim)
560 self._p.set_axes('xlabel',xlab)
561 self._p.set_axes('ylabel',ylab)
562 self._p.set_axes('title',tlab)
563 self._p.release()
564 if (not rcParams['plotter.gui']):
565 self._p.save(filename)
566 print_log()
567
568 def auto_fit(self, insitu=None, plot=False):
569 """
570 Return a scan where the function is applied to all rows for
571 all Beams/IFs/Pols.
572
573 """
574 from asap import scantable
575 if not isinstance(self.data, scantable) :
576 msg = "Data is not a scantable"
577 if rcParams['verbose']:
578 print msg
579 return
580 else:
581 raise TypeError(msg)
582 if insitu is None: insitu = rcParams['insitu']
583 if not insitu:
584 scan = self.data.copy()
585 else:
586 scan = self.data
587 rows = xrange(scan.nrow())
588 from asap import asaplog
589 asaplog.push("Fitting:")
590 for r in rows:
591 out = " Scan[%d] Beam[%d] IF[%d] Pol[%d] Cycle[%d]" % (scan.getscan(r),scan.getbeam(r),scan.getif(r),scan.getpol(r), scan.getcycle(r))
592 asaplog.push(out, False)
593 self.x = scan._getabcissa(r)
594 self.y = scan._getspectrum(r)
595 self.data = None
596 self.fit()
597 x = self.get_parameters()
598 if plot:
599 self.plot(residual=True)
600 x = raw_input("Accept fit ([y]/n): ")
601 if x.upper() == 'N':
602 continue
603 scan._setspectrum(self.fitter.getresidual(), r)
604 if plot:
605 self._p.unmap()
606 self._p = None
607 print_log()
608 return scan
609
Note: See TracBrowser for help on using the repository browser.