1 | import _asap
|
---|
2 | from asap import rcParams
|
---|
3 |
|
---|
4 | class fitter:
|
---|
5 | """
|
---|
6 | The fitting class for ASAP.
|
---|
7 | """
|
---|
8 | def _verbose(self, *args):
|
---|
9 | """
|
---|
10 | Set stdout output.
|
---|
11 | """
|
---|
12 | if type(args[0]) is bool:
|
---|
13 | self._vb = args[0]
|
---|
14 | return
|
---|
15 | elif len(args) == 0:
|
---|
16 | return self._vb
|
---|
17 |
|
---|
18 | def __init__(self):
|
---|
19 | """
|
---|
20 | Create a fitter object. No state is set.
|
---|
21 | """
|
---|
22 | self.fitter = _asap.fitter()
|
---|
23 | self.x = None
|
---|
24 | self.y = None
|
---|
25 | self.mask = None
|
---|
26 | self.fitfunc = None
|
---|
27 | self.fitfuncs = None
|
---|
28 | self.fitted = False
|
---|
29 | self.data = None
|
---|
30 | self.components = 0
|
---|
31 | self._fittedrow = 0
|
---|
32 | self._p = None
|
---|
33 | self._vb = True
|
---|
34 | self._selection = None
|
---|
35 |
|
---|
36 | def set_data(self, xdat, ydat, mask=None):
|
---|
37 | """
|
---|
38 | Set the absissa and ordinate for the fit. Also set the mask
|
---|
39 | indicationg valid points.
|
---|
40 | This can be used for data vectors retrieved from a scantable.
|
---|
41 | For scantable fitting use 'fitter.set_scan(scan, mask)'.
|
---|
42 | Parameters:
|
---|
43 | xdat: the abcissa values
|
---|
44 | ydat: the ordinate values
|
---|
45 | mask: an optional mask
|
---|
46 |
|
---|
47 | """
|
---|
48 | self.fitted = False
|
---|
49 | self.x = xdat
|
---|
50 | self.y = ydat
|
---|
51 | if mask == None:
|
---|
52 | from numarray import ones
|
---|
53 | self.mask = ones(len(xdat))
|
---|
54 | else:
|
---|
55 | self.mask = mask
|
---|
56 | return
|
---|
57 |
|
---|
58 | def set_scan(self, thescan=None, mask=None):
|
---|
59 | """
|
---|
60 | Set the 'data' (a scantable) of the fitter.
|
---|
61 | Parameters:
|
---|
62 | thescan: a scantable
|
---|
63 | mask: a msk retireved from the scantable
|
---|
64 | """
|
---|
65 | if not thescan:
|
---|
66 | print "Please give a correct scan"
|
---|
67 | self.fitted = False
|
---|
68 | self.data = thescan
|
---|
69 | if mask is None:
|
---|
70 | from numarray import ones
|
---|
71 | self.mask = ones(self.data.nchan())
|
---|
72 | else:
|
---|
73 | self.mask = mask
|
---|
74 | return
|
---|
75 |
|
---|
76 | def set_function(self, **kwargs):
|
---|
77 | """
|
---|
78 | Set the function to be fit.
|
---|
79 | Parameters:
|
---|
80 | poly: use a polynomial of the order given
|
---|
81 | gauss: fit the number of gaussian specified
|
---|
82 | Example:
|
---|
83 | fitter.set_function(gauss=2) # will fit two gaussians
|
---|
84 | fitter.set_function(poly=3) # will fit a 3rd order polynomial
|
---|
85 | """
|
---|
86 | #default poly order 0
|
---|
87 | n=0
|
---|
88 | if kwargs.has_key('poly'):
|
---|
89 | self.fitfunc = 'poly'
|
---|
90 | n = kwargs.get('poly')
|
---|
91 | self.components = [n]
|
---|
92 | elif kwargs.has_key('gauss'):
|
---|
93 | n = kwargs.get('gauss')
|
---|
94 | self.fitfunc = 'gauss'
|
---|
95 | self.fitfuncs = [ 'gauss' for i in range(n) ]
|
---|
96 | self.components = [ 3 for i in range(n) ]
|
---|
97 | else:
|
---|
98 | print "Invalid function type."
|
---|
99 | return
|
---|
100 | self.fitter.setexpression(self.fitfunc,n)
|
---|
101 | return
|
---|
102 |
|
---|
103 | def fit(self, row=0):
|
---|
104 | """
|
---|
105 | Execute the actual fitting process. All the state has to be set.
|
---|
106 | Parameters:
|
---|
107 | none
|
---|
108 | Example:
|
---|
109 | s = scantable('myscan.asap')
|
---|
110 | s.set_cursor(thepol=1) # select second pol
|
---|
111 | f = fitter()
|
---|
112 | f.set_scan(s)
|
---|
113 | f.set_function(poly=0)
|
---|
114 | f.fit(row=0) # fit first row
|
---|
115 | """
|
---|
116 | if ((self.x is None or self.y is None) and self.data is None) \
|
---|
117 | or self.fitfunc is None:
|
---|
118 | print "Fitter not yet initialised. Please set data & fit function"
|
---|
119 | return
|
---|
120 | else:
|
---|
121 | if self.data is not None:
|
---|
122 | self.x = self.data._getabcissa(row)
|
---|
123 | self.y = self.data._getspectrum(row)
|
---|
124 | print "Fitting:"
|
---|
125 | vb = self.data._vb
|
---|
126 | self.data._vb = True
|
---|
127 | self.selection = self.data.get_cursor()
|
---|
128 | self.data._vb = vb
|
---|
129 | self.fitter.setdata(self.x, self.y, self.mask)
|
---|
130 | if self.fitfunc == 'gauss':
|
---|
131 | ps = self.fitter.getparameters()
|
---|
132 | if len(ps) == 0:
|
---|
133 | self.fitter.estimate()
|
---|
134 | self.fitter.fit()
|
---|
135 | self._fittedrow = row
|
---|
136 | self.fitted = True
|
---|
137 | return
|
---|
138 |
|
---|
139 | def store_fit(self):
|
---|
140 | if self.fitted and self.data is not None:
|
---|
141 | pars = list(self.fitter.getparameters())
|
---|
142 | fixed = list(self.fitter.getfixedparameters())
|
---|
143 | self.data._addfit(self._fittedrow, pars, fixed,
|
---|
144 | self.fitfuncs, self.components)
|
---|
145 |
|
---|
146 | def set_parameters(self, params, fixed=None, component=None):
|
---|
147 | if self.fitfunc is None:
|
---|
148 | print "Please specify a fitting function first."
|
---|
149 | return
|
---|
150 | if self.fitfunc == "gauss" and component is not None:
|
---|
151 | if not self.fitted:
|
---|
152 | from numarray import zeros
|
---|
153 | pars = list(zeros(len(self.components)*3))
|
---|
154 | fxd = list(zeros(len(pars)))
|
---|
155 | else:
|
---|
156 | pars = list(self.fitter.getparameters())
|
---|
157 | fxd = list(self.fitter.getfixedparameters())
|
---|
158 | i = 3*component
|
---|
159 | pars[i:i+3] = params
|
---|
160 | fxd[i:i+3] = fixed
|
---|
161 | params = pars
|
---|
162 | fixed = fxd
|
---|
163 | self.fitter.setparameters(params)
|
---|
164 | if fixed is not None:
|
---|
165 | self.fitter.setfixedparameters(fixed)
|
---|
166 | return
|
---|
167 |
|
---|
168 | def set_gauss_parameters(self, peak, centre, fhwm,
|
---|
169 | peakfixed=False, centerfixed=False,
|
---|
170 | fhwmfixed=False,
|
---|
171 | component=0):
|
---|
172 | """
|
---|
173 | Set the Parameters of a 'Gaussian' component, set with set_function.
|
---|
174 | Parameters:
|
---|
175 | component: The number of the component (Default is the
|
---|
176 | first component.
|
---|
177 | peak, centre, fhwm: The gaussian parameters
|
---|
178 | peakfixed,
|
---|
179 | centerfixed,
|
---|
180 | fhwmfixed: Optional parameters to indicate if
|
---|
181 | the paramters should be held fixed during
|
---|
182 | the fitting process. The default is to keep
|
---|
183 | all parameters flexible.
|
---|
184 | """
|
---|
185 | if self.fitfunc != "gauss":
|
---|
186 | print "Function only operates on Gaussian components."
|
---|
187 | return
|
---|
188 | if 0 <= component < len(self.components):
|
---|
189 | self.set_parameters([peak, centre, fhwm],
|
---|
190 | [peakfixed, centerfixed, fhwmfixed],
|
---|
191 | component)
|
---|
192 | else:
|
---|
193 | print "Please select a valid component."
|
---|
194 | return
|
---|
195 |
|
---|
196 | def get_parameters(self, component=None):
|
---|
197 | """
|
---|
198 | Return the fit paramters.
|
---|
199 |
|
---|
200 | """
|
---|
201 | if not self.fitted:
|
---|
202 | print "Not yet fitted."
|
---|
203 | pars = list(self.fitter.getparameters())
|
---|
204 | fixed = list(self.fitter.getfixedparameters())
|
---|
205 | if component is not None:
|
---|
206 | if self.fitfunc == "gauss":
|
---|
207 | i = 3*component
|
---|
208 | cpars = pars[i:i+3]
|
---|
209 | cfixed = fixed[i:i+3]
|
---|
210 | else:
|
---|
211 | cpars = pars
|
---|
212 | cfixed = fixed
|
---|
213 | else:
|
---|
214 | cpars = pars
|
---|
215 | cfixed = fixed
|
---|
216 | fpars = self._format_pars(cpars, cfixed)
|
---|
217 | if self._vb:
|
---|
218 | print fpars
|
---|
219 | return cpars, cfixed, fpars
|
---|
220 |
|
---|
221 | def _format_pars(self, pars, fixed):
|
---|
222 | out = ''
|
---|
223 | if self.fitfunc == 'poly':
|
---|
224 | c = 0
|
---|
225 | for i in range(len(pars)):
|
---|
226 | fix = ""
|
---|
227 | if fixed[i]: fix = "(fixed)"
|
---|
228 | out += ' p%d%s= %3.3f,' % (c,fix,pars[i])
|
---|
229 | c+=1
|
---|
230 | out = out[:-1] # remove trailing ','
|
---|
231 | elif self.fitfunc == 'gauss':
|
---|
232 | i = 0
|
---|
233 | c = 0
|
---|
234 | aunit = ''
|
---|
235 | ounit = ''
|
---|
236 | if self.data:
|
---|
237 | aunit = self.data.get_unit()
|
---|
238 | ounit = self.data.get_fluxunit()
|
---|
239 | while i < len(pars):
|
---|
240 | out += ' %d: 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)
|
---|
241 | c+=1
|
---|
242 | i+=3
|
---|
243 | return out
|
---|
244 |
|
---|
245 | def get_estimate(self):
|
---|
246 | """
|
---|
247 | Return the parameter estimates (for non-linear functions).
|
---|
248 | """
|
---|
249 | pars = self.fitter.getestimate()
|
---|
250 | if self._vb:
|
---|
251 | print self._format_pars(pars)
|
---|
252 | return pars
|
---|
253 |
|
---|
254 |
|
---|
255 | def get_residual(self):
|
---|
256 | """
|
---|
257 | Return the residual of the fit.
|
---|
258 | """
|
---|
259 | if not self.fitted:
|
---|
260 | print "Not yet fitted."
|
---|
261 | return self.fitter.getresidual()
|
---|
262 |
|
---|
263 | def get_chi2(self):
|
---|
264 | """
|
---|
265 | Return chi^2.
|
---|
266 | """
|
---|
267 | if not self.fitted:
|
---|
268 | print "Not yet fitted."
|
---|
269 | ch2 = self.fitter.getchi2()
|
---|
270 | if self._vb:
|
---|
271 | print 'Chi^2 = %3.3f' % (ch2)
|
---|
272 | return ch2
|
---|
273 |
|
---|
274 | def get_fit(self):
|
---|
275 | """
|
---|
276 | Return the fitted ordinate values.
|
---|
277 | """
|
---|
278 | if not self.fitted:
|
---|
279 | print "Not yet fitted."
|
---|
280 | return self.fitter.getfit()
|
---|
281 |
|
---|
282 | def commit(self):
|
---|
283 | """
|
---|
284 | Return a new scan where the fits have been commited.
|
---|
285 | """
|
---|
286 | if not self.fitted:
|
---|
287 | print "Not yet fitted."
|
---|
288 | if self.data is not scantable:
|
---|
289 | print "Only works with scantables"
|
---|
290 | return
|
---|
291 | scan = self.data.copy()
|
---|
292 | scan._setspectrum(self.fitter.getresidual())
|
---|
293 |
|
---|
294 | def plot(self, residual=False, components=None, plotparms=False,
|
---|
295 | plotrange=None):
|
---|
296 | """
|
---|
297 | Plot the last fit.
|
---|
298 | Parameters:
|
---|
299 | residual: an optional parameter indicating if the residual
|
---|
300 | should be plotted (default 'False')
|
---|
301 | """
|
---|
302 | if not self.fitted:
|
---|
303 | return
|
---|
304 | if not self._p:
|
---|
305 | from asap.asaplot import ASAPlot
|
---|
306 | self._p = ASAPlot()
|
---|
307 | if self._p.is_dead:
|
---|
308 | from asap.asaplot import ASAPlot
|
---|
309 | self._p = ASAPlot()
|
---|
310 | self._p.clear()
|
---|
311 | self._p.set_panels()
|
---|
312 | self._p.palette(1)
|
---|
313 | tlab = 'Spectrum'
|
---|
314 | xlab = 'Abcissa'
|
---|
315 | m = ()
|
---|
316 | if self.data:
|
---|
317 | tlab = self.data._getsourcename(self._fittedrow)
|
---|
318 | xlab = self.data._getabcissalabel(self._fittedrow)
|
---|
319 | m = self.data._getmask(self._fittedrow)
|
---|
320 | ylab = r'Flux'
|
---|
321 |
|
---|
322 | colours = ["grey60","grey80","red","orange","purple","yellow","magenta", "cyan"]
|
---|
323 | self._p.palette(1,colours)
|
---|
324 | self._p.set_line(label='Spectrum')
|
---|
325 | self._p.plot(self.x, self.y, m)
|
---|
326 | if residual:
|
---|
327 | self._p.palette(2)
|
---|
328 | self._p.set_line(label='Residual')
|
---|
329 | self._p.plot(self.x, self.get_residual(), m)
|
---|
330 | self._p.palette(3)
|
---|
331 | if components is not None:
|
---|
332 | cs = components
|
---|
333 | if isinstance(components,int): cs = [components]
|
---|
334 | self._p.text(0.15,0.15,str(self.get_parameters()[2]),size=8)
|
---|
335 | n = len(self.components)
|
---|
336 | self._p.palette(4)
|
---|
337 | for c in cs:
|
---|
338 | if 0 <= c < n:
|
---|
339 | lab = self.fitfuncs[c]+str(c)
|
---|
340 | self._p.set_line(label=lab)
|
---|
341 | self._p.plot(self.x, self.fitter.evaluate(c), m)
|
---|
342 | elif c == -1:
|
---|
343 | self._p.palette(3)
|
---|
344 | self._p.set_line(label="Total Fit")
|
---|
345 | self._p.plot(self.x, self.get_fit(), m)
|
---|
346 | else:
|
---|
347 | self._p.palette(3)
|
---|
348 | self._p.set_line(label='Fit')
|
---|
349 | self._p.plot(self.x, self.get_fit(), m)
|
---|
350 | self._p.set_axes('xlabel',xlab)
|
---|
351 | self._p.set_axes('ylabel',ylab)
|
---|
352 | self._p.set_axes('title',tlab)
|
---|
353 | self._p.release()
|
---|
354 |
|
---|
355 | def auto_fit(self, insitu=None):
|
---|
356 | """
|
---|
357 | Return a scan where the function is applied to all rows for
|
---|
358 | all Beams/IFs/Pols.
|
---|
359 |
|
---|
360 | """
|
---|
361 | from asap import scantable
|
---|
362 | if not isinstance(self.data, scantable) :
|
---|
363 | print "Only works with scantables"
|
---|
364 | return
|
---|
365 | if insitu is None: insitu = rcParams['insitu']
|
---|
366 | if not insitu:
|
---|
367 | scan = self.data.copy()
|
---|
368 | else:
|
---|
369 | scan = self.data
|
---|
370 | vb = scan._vb
|
---|
371 | scan._vb = False
|
---|
372 | sel = scan.get_cursor()
|
---|
373 | rows = range(scan.nrow())
|
---|
374 | for i in range(scan.nbeam()):
|
---|
375 | scan.setbeam(i)
|
---|
376 | for j in range(scan.nif()):
|
---|
377 | scan.setif(j)
|
---|
378 | for k in range(scan.npol()):
|
---|
379 | scan.setpol(k)
|
---|
380 | if self._vb:
|
---|
381 | print "Fitting:"
|
---|
382 | print 'Beam[%d], IF[%d], Pol[%d]' % (i,j,k)
|
---|
383 | for iRow in rows:
|
---|
384 | self.x = scan._getabcissa(iRow)
|
---|
385 | self.y = scan._getspectrum(iRow)
|
---|
386 | self.data = None
|
---|
387 | self.fit()
|
---|
388 | x = self.get_parameters()
|
---|
389 | scan._setspectrum(self.fitter.getresidual(),iRow)
|
---|
390 | scan.set_cursor(sel[0],sel[1],sel[2])
|
---|
391 | scan._vb = vb
|
---|
392 | if not insitu:
|
---|
393 | return scan
|
---|