def __init__(self, resp, pmodel):
# check if response model has convolve method
if not hasattr(resp, "convolve"):
raise(PE.PyAAlgorithmFailure("Response model has no method 'convolve'.", \
where="ConvolutionModel::__init__"))
if not isinstance(pmodel, OneDFit):
raise(PE.PyAAlgorithmFailure("Physical model must be derived from OneDFit", \
where="ConvolutionModel::__init__"))
# Copy the input models
self.response = copy.deepcopy(resp)
self.pmodel = copy.deepcopy(pmodel)
if isinstance(self.response, OneDFit):
# The response is also a funcFit model and, therefore, could
# already contribute fitting parameters
#
# Change root name to "Response"
self.response.setRootName("Response")
# Ensure unique parameter names
self._combineRemapping(self.response, self.pmodel)
# List of parameters for the this (the convolved) model
parameters = list(self.response.parameters()) + list(
self.pmodel.parameters())
else:
# List of parameters for the this (the convolved) model
parameters = list(self.pmodel.parameters())
# Set up funcFit model and assign
OneDFit.__init__(self, parameters)
# Define root name
if self.pmodel.naming._root == "":
# Use default
rn = "model"
else:
rn = self.pmodel.naming._root
self.setRootName("Conv-" + rn)
def _readData(self):
"""
"""
# Determine number of planets in the csv file
r = csv.DictReader(self._fs.requestFile(self.dataFileName, 'rt',
gzip.open),
delimiter=',')
for nplanets, x in enumerate(r):
pass
# Reinitialize csv file
r = csv.DictReader(self._fs.requestFile(self.dataFileName, 'rt',
gzip.open),
delimiter=',')
# Determine data types for numpy recarray from columns
# and initialize
dtype = [(self._columns[x][0], self._columns[x][3])
for x in range(len(self._columns))]
self.data = np.recarray((nplanets + 1, ), dtype=dtype)
colnotfilled = [
self._columns[x][0] for x in six.iterkeys(self._columns)
]
for i, x in enumerate(r):
for k, v in six.iteritems(x):
# Remove hash and white spaces from column names
k = k.strip('#')
k = k.strip()
# Translate csv column name into internal column name
if k in self._ident:
key = self._ident[k]
else:
key = k
if len(v) == 0:
v = None
# Accept only expected fields
if not key in self.data.dtype.names:
continue
try:
colnotfilled.remove(key)
except ValueError:
# Ignoring already removed value
pass
self.data[key][i] = v
if len(colnotfilled) > 0:
PE.warn(
PE.PyAAlgorithmFailure(
"Not all columns could be filled with data. The following columns must not be used: "
+ ", ".join(colnotfilled),
where="ExoplanetEU",
solution=
"The format of the data base must be checked. Please consider issuing a bug report via github."
))
def selectWalkers(self, ws):
"""
Select walkers for emcee chains.
Parameters
----------
ws : list or array of integers
The walker to be considered in the analysis. Counting
starts at zero.
"""
self._selectedWalker = np.array(ws, dtype=np.int)
if self.dbtype == "emcee":
# Apply burn-in to individual walkers
self._loadEMCEEChain(burn=self.burn)
else:
raise(PE.PyAAlgorithmFailure("Walkers can only be selected for emcee data bases. Current data base type: " + str(self.dbtype)))
def fit(self,
m,
ds,
objf="chisqr",
initDelta=None,
maxIter=1e4,
callback=None,
nmCritLim=None):
"""
Carry out the model fit.
After the iteration, the `iterCount` attribute contains the
number of iterations. The `maxIterReached` attribute flag is
False, if the maximum number of iterations has not been reached
and True otherwise.
Parameters
----------
m : Instance of OneDFit
The model to be fitted.
ds : Instance of FufDS
The data.
objf : string
The objective function to be used. Possible
choices are "chisqr" (default), "sqrdiff", and
"cash79".
initDelta : dictionary, optional
A dictionary mapping parameter names to the
initial step width. This can be very useful, if
the starting values are zero or very small. The
here defined step will be added to the starting
value to construct the simplex.
maxIter : int, optional
The maximum number of iterations. The default is
10000.
nmCritLim : float, optional
Critical value for stopping criterion. The default is
1e-8.
callback : callable, optional
If not None, "callback" will be called with the
three parameters: number of iteration (int), current
best parameter set (array), and current simplex (array).
Returns
-------
Best-fit values : dictionary
Maps parameter name to the best-fit value.
"""
# Stopping criterion
if not nmCritLim is None:
self.nmCritLim = nmCritLim
# Number of free parameters
self._n = m.numberOfFreeParams()
# Set objective function
m.setObjectiveFunction(objf)
# Assign data object
m._fufDS = ds
# Names of free parameters (order guaranteed)
self._fpns = m.freeParamNames()
# Initial simplex
self._initSimplex(m, initDelta)
# MaxIter flag
self.maxIterReached = False
self.iterCount = 0
while (not self._stopCrit()) and (self.iterCount < maxIter):
self.iterCount += 1
self._step(m)
if callback is not None:
l = np.argmin(self._yi)
callback(self.iterCount, self._simplex[l, ::], self._simplex)
# Find the optimum parameter set
l = np.argmin(self._yi)
m.pars.setFreeParams(self._simplex[l, ::])
# Evaluate model so that model attribute holds the best match
m.evaluate(ds.x)
if self.iterCount == maxIter:
self.maxIterReached = True
PE.warn(
PE.PyAAlgorithmFailure(
"The maximum number of iterations has been reached.\n" +
"The fit may be inappropriate.",
where="NelderMead",
solution=[
"Increase number of iterations.",
"Change starting values.",
"Change algorithm parameters (e.g., alpha, beta, gamma)."
]))
# Return a dictionary with the best-bit parameters
return dict(zip(self._fpns, self._simplex[l, ::]))
def teffToColor_nop(self, band, teff, feH, stype="ms", noRaise=False):
"""
Converts effective temperature into color according to Eq. 1.
This method inverts Eq. 1. Note that the equation is parabolic
in the color (i.e., X). Therefore, there are two solutions of
which the one falling within the validity ranges specified in
Tables 4 and 5 of RM05 is selected. If none or both of the
solutions are valid, an exception is raised.
Parameters
----------
band : string
Band identifier.
teff : float
Effective temperature in K.
feH : float
Metallicity
stype : string, {ms, g}
Type of star (main sequence or giant).
noRaise : boolean, optional
If True, no exceptions will be raised, but warnings
will be given Both candidate solutions will be
returned in this case.
Returns
-------
X : float
Color in the specified band.
"""
self._checkBand(band)
self._checkST(stype)
if stype == "ms":
dat = self._tab2
datp = self._tab4
else:
dat = self._tab3
datp = self._tab5
# Use warnings of 'noRaise' is True
if noRaise:
raiser = PE.warn
else:
def throw(x):
raise (x)
raiser = throw
# Row of coefficients in the table.
j = self._bands.index(band)
c = dat[j, 0] + dat[j, 4] * feH + dat[j, 5] * feH**2 - 5040. / teff
a13p = dat[j, 1] + dat[j, 3] * feH
a2 = dat[j, 2]
sq = np.sqrt(a13p**2 - 4. * a2 * c)
X = [(-a13p + sq) / (2. * a2), (-a13p - sq) / (2. * a2)]
# Get validity ranges to decide, which solution is correct
mj = j * 4 + self._resolveMetallicityIndex(feH)
xmin = datp[mj, 1]
xmax = datp[mj, 2]
xin = [False] * 2
for i in smo.range(2):
if X[i] >= xmin and X[i] <= xmax:
xin[i] = True
if sum(xin) == 2:
raiser(PE.PyAAlgorithmFailure("No unique solution in inversion of teff-color relation.", \
where="teffToColor_nop",\
solution="Consider carefully(!) using 'noRaise'."))
# Only relevant if 'noRaise' is True
return tuple(X)
if sum(xin) == 0:
raiser(PE.PyAAlgorithmFailure("No valid solution in inversion of teff-color relation.\n" + \
"Band: " + str(band) + ", Range of validity: %.4e - %.4e" % (xmin, xmax) + \
", candidate solutions: %.4e and %.4e" % tuple(X) + ", Input Teff: %5.1f" % teff, \
where="teffToColor_nop",\
solution="Consider carefully(!) using 'noRaise'."))
# Only relevant if 'noRaise' is True
return tuple(X)
if xin[0]:
return X[0]
else:
return X[1]
