def draw(self, equation):
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary['IndependentData'][0]
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary['IndependentData'][1]
if not self.parent.z_data:
self.parent.z_data = equation.dataCache.allDataCacheDictionary['DependentData']
from mpl_toolkits.mplot3d import Axes3D # 3D apecific
from matplotlib import cm # to colormap from blue to red
if not self.parent.Z:
xModel = numpy.linspace(min(self.parent.x_data), max(self.parent.x_data), 20)
yModel = numpy.linspace(min(self.parent.y_data), max(self.parent.y_data), 20)
self.parent.X, self.parent.Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([self.parent.X, self.parent.Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
self.parent.Z = equation.CalculateModelPredictions(equation.solvedCoefficients, equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
self.axes = self.figure.gca(projection='3d')
self.axes.plot_surface(self.parent.X, self.parent.Y, self.parent.Z, rstride=1, cstride=1, cmap=cm.coolwarm,
linewidth=1, antialiased=True)
self.axes.scatter(self.parent.x_data, self.parent.y_data, self.parent.z_data)
self.axes.set_title('Surface Plot (click-drag with mouse)') # add a title for surface plot
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
self.axes.set_zlabel('Z Data') # Z axis data label
def draw(self, equation):
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary['IndependentData'][0]
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary['IndependentData'][1]
if not self.parent.z_data:
self.parent.z_data = equation.dataCache.allDataCacheDictionary['DependentData']
if not self.parent.Z:
xModel = numpy.linspace(min(self.parent.x_data), max(self.parent.x_data), 20)
yModel = numpy.linspace(min(self.parent.y_data), max(self.parent.y_data), 20)
self.parent.X, self.parent.Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([self.parent.X, self.parent.Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
self.parent.Z = equation.CalculateModelPredictions(equation.solvedCoefficients, equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
self.axes.plot(self.parent.x_data, self.parent.y_data, 'o')
self.axes.set_title('Contour Plot') # add a title for contour plot
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
numberOfContourLines = 16
CS = matplotlib.pyplot.contour(self.parent.X, self.parent.Y, self.parent.Z, numberOfContourLines, colors='k')
matplotlib.pyplot.clabel(CS, inline=1, fontsize=10) # labels for contours
def SurfacePlot(dataObject, inFileName):
# raw data for scatterplot
x_data = dataObject.IndependentDataArray[0]
y_data = dataObject.IndependentDataArray[1]
z_data = dataObject.DependentDataArray
# create X, Y, Z mesh grid for surface plot
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = dataObject.equation.dataCache # temporarily store cache
dataObject.equation.dataCache = pyeq2.dataCache()
dataObject.equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([X, Y])
dataObject.equation.dataCache.FindOrCreateAllDataCache(dataObject.equation)
Z = dataObject.equation.CalculateModelPredictions(
dataObject.equation.solvedCoefficients,
dataObject.equation.dataCache.allDataCacheDictionary)
dataObject.equation.dataCache = tempcache # restore cache
# matplotlib specific code for the plots
fig = plt.figure(figsize=(float(dataObject.graphWidth) / 100.0,
float(dataObject.graphHeight) / 100.0),
dpi=100)
fig.patch.set_visible(False)
ax = fig.gca(projection='3d')
# create a surface plot using the X, Y, Z mesh data created above
ax.plot_surface(X,
Y,
Z,
rstride=1,
cstride=1,
cmap=cm.coolwarm,
linewidth=1,
antialiased=True)
# create a scatter plot of the raw data
if dataObject.dataPointSize3D == 0.0: # auto
ax.scatter(x_data, y_data, z_data)
else:
ax.scatter(x_data, y_data, z_data, s=dataObject.dataPointSize3D)
ax.set_title("Surface Plot") # add a title for surface plot
ax.set_xlabel(dataObject.IndependentDataName1) # X axis data label
ax.set_ylabel(dataObject.IndependentDataName2) # Y axis data label
ax.set_zlabel(dataObject.DependentDataName) # Z axis data label
if not inFileName:
return [fig, ax, plt] # for use in creating animations
else:
fig.savefig(inFileName[:-3] + 'png', format='png')
if not dataObject.pngOnlyFlag: # function finder results are png-only
fig.savefig(
inFileName[:-3] + 'svg',
format='svg',
)
plt.close('all')
def SaveModelScatterConfidence(in_filePath, in_equation, in_title, in_xAxisLabel, in_yAxisLabel):
# raw data
x_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = in_equation.dataCache.allDataCacheDictionary['DependentData']
# now create data for the fitted in_equation plot
xModel = numpy.linspace(min(x_data), max(x_data))
tempcache = in_equation.dataCache
in_equation.dataCache = pyeq2.dataCache()
in_equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([xModel, xModel])
in_equation.dataCache.FindOrCreateAllDataCache(in_equation)
yModel = in_equation.CalculateModelPredictions(in_equation.solvedCoefficients, in_equation.dataCache.allDataCacheDictionary)
in_equation.dataCache = tempcache
# now use matplotlib to create the PNG file
fig = plt.figure(figsize=(5, 4))
ax = fig.add_subplot(1,1,1)
# first the raw data as a scatter plot
ax.plot(x_data, y_data, 'D')
# now the model as a line plot
ax.plot(xModel, yModel)
# now calculate confidence intervals
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(x_data) # mean of x
n = in_equation.nobs # number of samples in the origional fit
t_value = scipy.stats.t.ppf(0.975, in_equation.df_e) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt((in_equation.sumOfSquaredErrors/in_equation.df_e)*(1.0/n + (numpy.power((xModel-mean_x),2.0)/
((numpy.sum(numpy.power(x_data,2)))-n*(numpy.power(mean_x,2.0))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yModel + abs(confs)
lower = yModel - abs(confs)
# mask off any numbers outside the existing plot limits
booleanMask = yModel > matplotlib.pyplot.ylim()[0]
booleanMask &= (yModel < matplotlib.pyplot.ylim()[1])
# color scheme improves visibility on black background lines or points
ax.plot(xModel[booleanMask], lower[booleanMask], linestyle='solid', color='white')
ax.plot(xModel[booleanMask], upper[booleanMask], linestyle='solid', color='white')
ax.plot(xModel[booleanMask], lower[booleanMask], linestyle='dashed', color='blue')
ax.plot(xModel[booleanMask], upper[booleanMask], linestyle='dashed', color='blue')
ax.set_title(in_title) # add a title
ax.set_xlabel(in_xAxisLabel) # X axis data label
ax.set_ylabel(in_yAxisLabel) # Y axis data label
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePath) # create PNG file
plt.close('all')
def SurfaceAndContourPlots(in_filePathSurface, in_filePathContour, in_equation,
in_surfaceTitle, in_contourTitle,
in_xAxisLabel, in_yAxisLabel, in_zAxisLabel):
# raw data
x_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][1]
z_data = in_equation.dataCache.allDataCacheDictionary['DependentData']
from mpl_toolkits.mplot3d import Axes3D # 3D apecific
from matplotlib import cm # to colormap from blue to red
fig = plt.figure()
ax = fig.gca(projection='3d')
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = in_equation.dataCache
in_equation.dataCache = pyeq2.dataCache()
in_equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([X, Y])
in_equation.dataCache.FindOrCreateAllDataCache(in_equation)
Z = in_equation.CalculateModelPredictions(in_equation.solvedCoefficients, in_equation.dataCache.allDataCacheDictionary)
in_equation.dataCache = tempcache
ax.plot_surface(X, Y, Z, rstride=1, cstride=1, cmap=cm.coolwarm,
linewidth=1, antialiased=True)
ax.scatter(x_data, y_data, z_data)
ax.set_title(in_surfaceTitle) # add a title for surface plot
ax.set_xlabel(in_xAxisLabel) # X axis data label
ax.set_ylabel(in_yAxisLabel) # Y axis data label
ax.set_zlabel(in_zAxisLabel) # Y axis data label
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePathSurface) # create PNG file
plt.close('all')
# contour plot here
fig = plt.figure(figsize=(5, 4))
ax = fig.add_subplot(1,1,1)
ax.plot(x_data, y_data, 'o', color='0.8', markersize=4) # draw these first so contour lines overlay. Color=number is grayscale
ax.set_title(in_contourTitle) # add a title for contour plot
ax.set_xlabel(in_xAxisLabel) # X data label
ax.set_ylabel(in_yAxisLabel) # Y data label
numberOfContourLines = 16
CS = plt.contour(X, Y, Z, numberOfContourLines, colors='k')
plt.clabel(CS, inline=1, fontsize=10) # labels for contours
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePathContour) # create PNG file
plt.close('all')
def draw(self, equation):
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary['DependentData']
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary['IndependentData'][0]
# now create data for the fitted equation plot
xModel = numpy.linspace(min(self.parent.x_data), max(self.parent.x_data))
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([xModel, xModel])
equation.dataCache.FindOrCreateAllDataCache(equation)
yModel = equation.CalculateModelPredictions(equation.solvedCoefficients, equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
# first the raw data as a scatter plot
self.axes.plot(self.parent.x_data, self.parent.y_data, 'D')
# now the model as a line plot
self.axes.plot(xModel, yModel)
# now calculate confidence intervals
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(self.parent.x_data) # mean of x
n = equation.nobs # number of samples in the origional fit
t_value = scipy.stats.t.ppf(0.975, equation.df_e) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt((equation.sumOfSquaredErrors/equation.df_e)*(1.0/n + (numpy.power((xModel-mean_x),2.0)/
((numpy.sum(numpy.power(self.parent.x_data,2)))-n*(numpy.power(mean_x,2.0))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yModel + abs(confs)
lower = yModel - abs(confs)
# mask off any numbers outside the existing plot limits
booleanMask = yModel > matplotlib.pyplot.ylim()[0]
booleanMask &= (yModel < matplotlib.pyplot.ylim()[1])
# color scheme improves visibility on black background lines or points
self.axes.plot(xModel[booleanMask], lower[booleanMask], linestyle='solid', color='white')
self.axes.plot(xModel[booleanMask], upper[booleanMask], linestyle='solid', color='white')
self.axes.plot(xModel[booleanMask], lower[booleanMask], linestyle='dashed', color='blue')
self.axes.plot(xModel[booleanMask], upper[booleanMask], linestyle='dashed', color='blue')
self.axes.set_title('Model With 95% Confidence Limits') # add a title
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
def draw(self, equation):
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary[
'IndependentData'][0]
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary[
'IndependentData'][1]
if not self.parent.z_data:
self.parent.z_data = equation.dataCache.allDataCacheDictionary[
'DependentData']
from mpl_toolkits.mplot3d import Axes3D # 3D apecific
from matplotlib import cm # to colormap from blue to red
if not self.parent.Z:
xModel = numpy.linspace(min(self.parent.x_data),
max(self.parent.x_data), 20)
yModel = numpy.linspace(min(self.parent.y_data),
max(self.parent.y_data), 20)
self.parent.X, self.parent.Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array(
[self.parent.X, self.parent.Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
self.parent.Z = equation.CalculateModelPredictions(
equation.solvedCoefficients,
equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
self.axes = self.figure.gca(projection='3d')
self.axes.plot_surface(self.parent.X,
self.parent.Y,
self.parent.Z,
rstride=1,
cstride=1,
cmap=cm.coolwarm,
linewidth=1,
antialiased=True)
self.axes.scatter(self.parent.x_data, self.parent.y_data,
self.parent.z_data)
self.axes.set_title('Surface Plot (click-drag with mouse)'
) # add a title for surface plot
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
self.axes.set_zlabel('Z Data') # Z axis data label
def SurfacePlot(dataObject, inFileName):
# raw data for scatterplot
x_data = dataObject.IndependentDataArray[0]
y_data = dataObject.IndependentDataArray[1]
z_data = dataObject.DependentDataArray
# create X, Y, Z mesh grid for surface plot
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = dataObject.equation.dataCache # temporarily store cache
dataObject.equation.dataCache = pyeq2.dataCache()
dataObject.equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([X, Y])
dataObject.equation.dataCache.FindOrCreateAllDataCache(dataObject.equation)
Z = dataObject.equation.CalculateModelPredictions(dataObject.equation.solvedCoefficients, dataObject.equation.dataCache.allDataCacheDictionary)
dataObject.equation.dataCache = tempcache # restore cache
# matplotlib specific code for the plots
fig = plt.figure(figsize=(float(dataObject.graphWidth) / 100.0, float(dataObject.graphHeight) / 100.0), dpi=100)
fig.patch.set_visible(False)
ax = fig.gca(projection='3d')
# create a surface plot using the X, Y, Z mesh data created above
ax.plot_surface(X, Y, Z, rstride=1, cstride=1, cmap=cm.coolwarm, linewidth=1, antialiased=True)
# create a scatter plot of the raw data
if dataObject.dataPointSize3D == 0.0: # auto
ax.scatter(x_data, y_data, z_data)
else:
ax.scatter(x_data, y_data, z_data, s=dataObject.dataPointSize3D)
ax.set_title("Surface Plot") # add a title for surface plot
ax.set_xlabel(dataObject.IndependentDataName1) # X axis data label
ax.set_ylabel(dataObject.IndependentDataName2) # Y axis data label
ax.set_zlabel(dataObject.DependentDataName) # Z axis data label
if not inFileName:
return [fig, ax, plt] # for use in creating animations
else:
fig.savefig(inFileName[:-3] + 'png', format = 'png')
if not dataObject.pngOnlyFlag: # function finder results are png-only
fig.savefig(inFileName[:-3] + 'svg', format = 'svg', )
plt.close('all')
def draw(self, equation):
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary[
'IndependentData'][0]
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary[
'IndependentData'][1]
if not self.parent.z_data:
self.parent.z_data = equation.dataCache.allDataCacheDictionary[
'DependentData']
if not self.parent.Z:
xModel = numpy.linspace(min(self.parent.x_data),
max(self.parent.x_data), 20)
yModel = numpy.linspace(min(self.parent.y_data),
max(self.parent.y_data), 20)
self.parent.X, self.parent.Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array(
[self.parent.X, self.parent.Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
self.parent.Z = equation.CalculateModelPredictions(
equation.solvedCoefficients,
equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
self.axes.plot(self.parent.x_data, self.parent.y_data, 'o')
self.axes.set_title('Contour Plot') # add a title for contour plot
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
numberOfContourLines = 16
CS = matplotlib.pyplot.contour(self.parent.X,
self.parent.Y,
self.parent.Z,
numberOfContourLines,
colors='k')
matplotlib.pyplot.clabel(CS, inline=1,
fontsize=10) # labels for contours
def SaveModelScatterConfidence(in_filePath, in_equation, in_title,
in_xAxisLabel, in_yAxisLabel):
# raw data
x_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = in_equation.dataCache.allDataCacheDictionary['DependentData']
# now create data for the fitted in_equation plot
xModel = numpy.linspace(min(x_data), max(x_data))
tempcache = in_equation.dataCache
in_equation.dataCache = pyeq2.dataCache()
in_equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([xModel, xModel])
in_equation.dataCache.FindOrCreateAllDataCache(in_equation)
yModel = in_equation.CalculateModelPredictions(
in_equation.solvedCoefficients,
in_equation.dataCache.allDataCacheDictionary)
in_equation.dataCache = tempcache
# now use matplotlib to create the PNG file
fig = plt.figure(figsize=(5, 4))
ax = fig.add_subplot(1, 1, 1)
# first the raw data as a scatter plot
ax.plot(x_data, y_data, 'D')
# now the model as a line plot
ax.plot(xModel, yModel)
# now calculate confidence intervals
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(x_data) # mean of x
n = in_equation.nobs # number of samples in the origional fit
t_value = scipy.stats.t.ppf(
0.975, in_equation.df_e
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(in_equation.sumOfSquaredErrors / in_equation.df_e) *
(1.0 / n + (numpy.power((xModel - mean_x), 2.0) /
((numpy.sum(numpy.power(x_data, 2))) - n *
(numpy.power(mean_x, 2.0))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yModel + abs(confs)
lower = yModel - abs(confs)
# mask off any numbers outside the existing plot limits
booleanMask = yModel > matplotlib.pyplot.ylim()[0]
booleanMask &= (yModel < matplotlib.pyplot.ylim()[1])
# color scheme improves visibility on black background lines or points
ax.plot(xModel[booleanMask],
lower[booleanMask],
linestyle='solid',
color='white')
ax.plot(xModel[booleanMask],
upper[booleanMask],
linestyle='solid',
color='white')
ax.plot(xModel[booleanMask],
lower[booleanMask],
linestyle='dashed',
color='blue')
ax.plot(xModel[booleanMask],
upper[booleanMask],
linestyle='dashed',
color='blue')
ax.set_title(in_title) # add a title
ax.set_xlabel(in_xAxisLabel) # X axis data label
ax.set_ylabel(in_yAxisLabel) # Y axis data label
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePath) # create PNG file
plt.close('all')
def ParallelFittingFunction(rawData, fittingTargetText, smoothnessControl,
modulus, modulusRemainder):
processID = str(os.getpid())
# this function yields a single item to inspect after completion
bestResult = []
# we are using the same data set repeatedly, so create a cache external to the equations
externalCache = pyeq2.dataCache()
reducedDataCache = {}
equationCountForModulus = 0
##########################
# add named equations here
print
print 'Process ID', processID, 'fitting named equations:'
for submodule in inspect.getmembers(pyeq2.Models_2D):
if inspect.ismodule(submodule[1]):
for equationClass in inspect.getmembers(submodule[1]):
if inspect.isclass(equationClass[1]):
# special classes
if equationClass[1].splineFlag or \
equationClass[1].userSelectablePolynomialFlag or \
equationClass[1].userSelectablePolyfunctionalFlag or \
equationClass[1].userSelectableRationalFlag or \
equationClass[1].userDefinedFunctionFlag:
continue
for extendedVersion in ['Default', 'Offset']:
if (extendedVersion == 'Offset') and (
equationClass[1].autoGenerateOffsetForm
== False):
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = equationClass[1](fittingTargetText,
extendedVersion)
if len(equationInstance.GetCoefficientDesignators()
) > smoothnessControl:
continue
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService(
).ConvertAndSortColumnarASCII(
rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(
equationInstance)
if reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[
equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance,
bestResult, True,
processID)
if result:
bestResult = result
if not reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
reducedDataCache[
equationInstance.
numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
##########################
# fit polyfunctionals here
print
print 'Process ID', processID, 'fitting polyfunctionals:'
equationCount = 0
maxPolyfunctionalCoefficients = 4 # this value was chosen to make this example more convenient
polyfunctionalEquationList = pyeq2.PolyFunctions.GenerateListForPolyfunctionals_2D(
)
functionIndexList = range(len(polyfunctionalEquationList)
) # make a list of function indices to permute
for coeffCount in range(1, maxPolyfunctionalCoefficients + 1):
functionCombinations = UniqueCombinations(functionIndexList,
coeffCount)
for functionCombination in functionCombinations:
if len(functionCombination) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Polyfunctional.UserSelectablePolyfunctional(
fittingTargetText, 'Default', functionCombination,
polyfunctionalEquationList)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(
rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(
equationInstance)
if reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[
equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
reducedDataCache[
equationInstance.
numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
equationCount += 1
if (equationCount % 50) == 0:
print ' Process ID', processID, 'fitted', equationCount, 'equations...'
######################
# fit user-selectable polynomials here
print
print 'Process ID', processID, 'fitting user-selectable polynomials:'
maxPolynomialOrderX = 5 # this value was chosen to make this example more convenient
for polynomialOrderX in range(maxPolynomialOrderX + 1):
if (polynomialOrderX + 1) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Polynomial.UserSelectablePolynomial(
fittingTargetText, 'Default', polynomialOrderX)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(
rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(
equationInstance)
if reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[
equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
reducedDataCache[
equationInstance.
numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
######################
# fit user-selectable rationals here
print
print 'Process ID', processID, 'fitting user-selectable rationals:'
equationCount = 0
maxCoeffs = smoothnessControl # arbitrary choice of maximum total coefficients for this example
functionList = pyeq2.PolyFunctions.GenerateListForRationals_2D()
functionIndexList = range(
len(functionList)) # make a list of function indices
for numeratorCoeffCount in range(1, maxCoeffs):
numeratorComboList = UniqueCombinations(functionIndexList,
numeratorCoeffCount)
for numeratorCombo in numeratorComboList:
for denominatorCoeffCount in range(1, maxCoeffs):
denominatorComboList = UniqueCombinations2(
functionIndexList, denominatorCoeffCount)
for denominatorCombo in denominatorComboList:
for extendedVersion in ['Default', 'Offset']:
extraCoeffs = 0
if extendedVersion == 'Offset':
extraCoeffs = 1
if (len(numeratorCombo) + len(denominatorCombo) +
extraCoeffs) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Rational.UserSelectableRational(
fittingTargetText, extendedVersion, numeratorCombo,
denominatorCombo, functionList)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService(
).ConvertAndSortColumnarASCII(
rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(
equationInstance)
if reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[
equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance,
bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(
equationInstance.numberOfReducedDataPoints):
reducedDataCache[
equationInstance.
numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
equationCount += 1
if (equationCount % 5) == 0:
print ' ', 'Process ID', processID + ',', equationCount, 'rationals, current flags:', equationInstance.rationalNumeratorFlags, equationInstance.rationalDenominatorFlags,
if extendedVersion == 'Offset':
print 'with offset'
else:
print
print 'Process ID', processID, 'has completed'
return bestResult
def __init__(self, inFittingTarget = 'SSQABS', inExtendedVersionName = 'Default'):
if inExtendedVersionName == '':
inExtendedVersionName = 'Default'
if inFittingTarget not in self.fittingTargetDictionary.keys():
raise Exception, str(inFittingTarget) + ' is not in the IModel class fitting target dictionary.'
self.extendedVersionHandler = eval('pyeq2.ExtendedVersionHandlers.ExtendedVersionHandler_' + inExtendedVersionName.replace(' ', '') + '.ExtendedVersionHandler_' + inExtendedVersionName.replace(' ', '') + '()')
self.dataCache = pyeq2.dataCache()
self.upperCoefficientBounds = []
self.lowerCoefficientBounds = []
self.estimatedCoefficients = []
self.fixedCoefficients = []
self.solvedCoefficients = []
self.polyfunctional2DFlags = []
self.polyfunctional3DFlags = []
self.xPolynomialOrder = None
self.yPolynomialOrder = None
self.rationalNumeratorFlags = []
self.rationalDenominatorFlags = []
self.fittingTarget = inFittingTarget
self.deEstimatedCoefficients = []
self.independentData1CannotContainZeroFlag = False
self.independentData1CannotContainPositiveFlag = False
self.independentData1CannotContainNegativeFlag = False
self.independentData2CannotContainZeroFlag = False
self.independentData2CannotContainPositiveFlag = False
self.independentData2CannotContainNegativeFlag = False
try:
if self._dimensionality == 2:
self.exampleData = '''
X Y
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
9.769 7.068
9.861 7.104
'''
else:
self.exampleData = '''
X Y Z
3.017 2.175 0.320
2.822 2.624 0.629
2.632 2.839 0.950
2.287 3.030 1.574
2.207 3.057 1.725
2.048 3.098 2.035
1.963 3.115 2.204
1.784 3.144 2.570
1.712 3.153 2.721
2.972 2.106 0.313
2.719 2.542 0.643
2.495 2.721 0.956
2.070 2.878 1.597
1.969 2.899 1.758
1.768 2.929 2.088
1.677 2.939 2.240
1.479 2.957 2.583
1.387 2.963 2.744
2.843 1.984 0.315
2.485 2.320 0.639
2.163 2.444 0.954
1.687 2.525 1.459
1.408 2.547 1.775
1.279 2.554 1.927
1.016 2.564 2.243
0.742 2.568 2.581
0.607 2.571 2.753
'''
except:
pass
def ContourPlot(in_DataObject, in_FileNameAndPath):
gridResolution = (in_DataObject.graphWidth +
in_DataObject.graphHeight) / 20
gxmin = in_DataObject.gxmin
gxmax = in_DataObject.gxmax
gymin = in_DataObject.gymin
gymax = in_DataObject.gymax
if in_DataObject.equation.independentData1CannotContainNegativeFlag and gxmin < 0.0:
gxmin = 0.0
if in_DataObject.equation.independentData1CannotContainZeroFlag and gxmin == 0.0:
gxmin = 1.0E-300
if in_DataObject.equation.independentData1CannotContainPositiveFlag and gxmax > 0.0:
gxmax = 0.0
if in_DataObject.equation.independentData1CannotContainZeroFlag and gxmax == 0.0:
gxmax = 1.0E-300
if in_DataObject.equation.independentData2CannotContainNegativeFlag and gymin < 0.0:
gymin = 0.0
if in_DataObject.equation.independentData2CannotContainZeroFlag and gymin == 0.0:
gymin = 1.0E-300
if in_DataObject.equation.independentData2CannotContainPositiveFlag and gymax > 0.0:
gymax = 0.0
if in_DataObject.equation.independentData2CannotContainZeroFlag and gymax == 0.0:
gymax = 1.0E-300
deltax = (gxmax - gxmin) / float(gridResolution)
deltay = (gymax - gymin) / float(gridResolution)
xRange = numpy.arange(gxmin, gxmax + deltax, deltax)
yRange = numpy.arange(gymin, gymax + deltay / 2.0, deltay)
minZ = min(in_DataObject.DependentDataArray)
maxZ = max(in_DataObject.DependentDataArray)
X, Y = numpy.meshgrid(xRange, yRange)
boundingRectangle = matplotlib.patches.Rectangle([gxmin, gymin],
gxmax - gxmin,
gymax - gymin,
facecolor=(0.975, 0.975,
0.975),
edgecolor=(0.9, 0.9, 0.9))
Z = []
tempDataCache = in_DataObject.equation.dataCache
for i in range(len(X)):
in_DataObject.equation.dataCache = pyeq2.dataCache()
in_DataObject.equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([X[i], Y[i]])
in_DataObject.equation.dataCache.FindOrCreateAllDataCache(
in_DataObject.equation)
Z.append(
in_DataObject.equation.CalculateModelPredictions(
in_DataObject.equation.solvedCoefficients,
in_DataObject.equation.dataCache.allDataCacheDictionary))
in_DataObject.equation.dataCache = tempDataCache
Z = numpy.array(Z)
Z = numpy.clip(Z, minZ, maxZ)
tempData = [
in_DataObject.IndependentDataArray[0],
in_DataObject.IndependentDataArray[1], in_DataObject.DependentDataArray
]
ContourPlot_NoDataObject(
X, Y, Z, tempData, in_FileNameAndPath,
in_DataObject.IndependentDataName1, in_DataObject.IndependentDataName2,
in_DataObject.graphWidth, in_DataObject.graphHeight, 'UseOffset_ON',
'ScientificNotation_X_AUTO', 'ScientificNotation_Y_AUTO',
in_DataObject.pngOnlyFlag, boundingRectangle)
plt.close()
def __init__(self,
inFittingTarget='SSQABS',
inExtendedVersionName='Default'):
if inExtendedVersionName == '':
inExtendedVersionName = 'Default'
if inFittingTarget not in list(self.fittingTargetDictionary.keys()):
raise Exception(
str(inFittingTarget) +
' is not in the IModel class fitting target dictionary.')
self.fittingTarget = inFittingTarget
inExtendedVersionName = inExtendedVersionName.replace(' ', '')
if inExtendedVersionName not in pyeq2.ExtendedVersionHandlers.extendedVersionHandlerNameList:
raise Exception(
inExtendedVersionName +
' is not in the list of extended version handler names.')
allowedExtendedVersion = True
if (-1 != inExtendedVersionName.find('Offset')) and (
self.autoGenerateOffsetForm == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Reciprocal')) and (
self.autoGenerateReciprocalForm == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Inverse')) and (
self.autoGenerateInverseForms == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Growth')) and (
self.autoGenerateGrowthAndDecayForms == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Decay')) and (
self.autoGenerateGrowthAndDecayForms == False):
allowedExtendedVersion = False
if allowedExtendedVersion == False:
raise Exception(
'This equation does not allow an extended version named "' +
inExtendedVersionName + '".')
self.extendedVersionHandler = eval(
'pyeq2.ExtendedVersionHandlers.ExtendedVersionHandler_' +
inExtendedVersionName + '.ExtendedVersionHandler_' +
inExtendedVersionName + '()')
self.dataCache = pyeq2.dataCache()
self.upperCoefficientBounds = []
self.lowerCoefficientBounds = []
self.estimatedCoefficients = []
self.fixedCoefficients = []
self.solvedCoefficients = []
self.polyfunctional2DFlags = []
self.polyfunctional3DFlags = []
self.xPolynomialOrder = None
self.yPolynomialOrder = None
self.rationalNumeratorFlags = []
self.rationalDenominatorFlags = []
self.deEstimatedCoefficients = []
try:
if self._dimensionality == 2:
self.exampleData = '''
X Y
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
9.769 7.068
9.861 7.104
'''
else:
self.exampleData = '''
X Y Z
3.017 2.175 0.320
2.822 2.624 0.629
2.632 2.839 0.950
2.287 3.030 1.574
2.207 3.057 1.725
2.048 3.098 2.035
1.963 3.115 2.204
1.784 3.144 2.570
1.712 3.153 2.721
2.972 2.106 0.313
2.719 2.542 0.643
2.495 2.721 0.956
2.070 2.878 1.597
1.969 2.899 1.758
1.768 2.929 2.088
1.677 2.939 2.240
1.479 2.957 2.583
1.387 2.963 2.744
2.843 1.984 0.315
2.485 2.320 0.639
2.163 2.444 0.954
1.687 2.525 1.459
1.408 2.547 1.775
1.279 2.554 1.927
1.016 2.564 2.243
0.742 2.568 2.581
0.607 2.571 2.753
'''
except:
pass
def __init__(self,
inFittingTarget='SSQABS',
inExtendedVersionName='Default'):
self.dataCache = pyeq2.dataCache()
self._dimensionality = 3
def ScatterPlotWithOptionalModel_NoDataObject(
in_DataToPlot, in_FileNameAndPath, in_DataNameX, in_DataNameY,
in_WidthInPixels, in_HeightInPixels, in_Equation, in_UseOffsetIfNeeded,
in_ReverseXY, in_X_UseScientificNotationIfNeeded,
in_Y_UseScientificNotationIfNeeded, in_GraphBounds, in_LogY, in_LogX,
inPNGOnlyFlag, inConfidenceIntervalsFlag):
# decode ends of strings ('XYZ_ON', 'XYZ_OFF', 'XYZ_AUTO', etc.) to boolean values
scientificNotationX = DetermineScientificNotationFromString(
in_DataToPlot[0], in_X_UseScientificNotationIfNeeded)
scientificNotationY = DetermineScientificNotationFromString(
in_DataToPlot[1], in_Y_UseScientificNotationIfNeeded)
useOffsetIfNeeded = DetermineOnOrOffFromString(in_UseOffsetIfNeeded)
reverseXY = DetermineOnOrOffFromString(in_ReverseXY)
if in_Equation: # make model data for plotting, clipping at boundaries
lowerXbound = in_GraphBounds[0]
upperXbound = in_GraphBounds[1]
if in_Equation.independentData1CannotContainNegativeFlag and lowerXbound < 0.0:
lowerXbound = 0.0
if in_Equation.independentData1CannotContainZeroFlag and lowerXbound == 0.0:
lowerXbound = 1.0E-300
if in_Equation.independentData1CannotContainPositiveFlag and upperXbound > 0.0:
upperXbound = 0.0
if in_Equation.independentData1CannotContainZeroFlag and upperXbound == 0.0:
upperXbound = 1.0E-300
xRange = numpy.arange(
lowerXbound, upperXbound, (upperXbound - lowerXbound) /
(20.0 * float(in_WidthInPixels + in_HeightInPixels))
) # make this 'reverse-xy-independent'
tempDataCache = in_Equation.dataCache
in_Equation.dataCache = pyeq2.dataCache()
in_Equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([xRange, xRange])
in_Equation.dataCache.FindOrCreateAllDataCache(in_Equation)
yRange = in_Equation.CalculateModelPredictions(
in_Equation.solvedCoefficients,
in_Equation.dataCache.allDataCacheDictionary)
in_Equation.dataCache = tempDataCache
if reverseXY:
fig, ax = CommonPlottingCode(in_WidthInPixels, in_HeightInPixels,
in_DataNameX, in_DataNameY,
useOffsetIfNeeded, scientificNotationY,
scientificNotationX, 0.0, 0.0, 1.0, 1.0)
ax.plot(numpy.array([in_GraphBounds[2], in_GraphBounds[3]]),
numpy.array([in_GraphBounds[0], in_GraphBounds[1]
])) # first ax.plot() is only with extents
newLeft, newBottom, newRight, newTop, numberOfMajor_X_TickMarks = YieldNewExtentsAndNumberOfMajor_X_TickMarks(
fig, ax, in_WidthInPixels, in_HeightInPixels, scientificNotationY
or useOffsetIfNeeded)
fig, ax = CommonPlottingCode(in_WidthInPixels, in_HeightInPixels,
in_DataNameX, in_DataNameY,
useOffsetIfNeeded, scientificNotationY,
scientificNotationX, newLeft, newBottom,
newRight, newTop)
if in_LogY == 'LOG' and in_LogX == 'LOG':
loglinplot = ax.loglog
elif in_LogY == 'LIN' and in_LogX == 'LOG':
loglinplot = ax.semilogx
elif in_LogY == 'LOG' and in_LogX == 'LIN':
loglinplot = ax.semilogy
else:
loglinplot = ax.plot
if len(ax.get_xticklabels()) > numberOfMajor_X_TickMarks:
ax.xaxis.set_major_locator(
matplotlib.ticker.MaxNLocator(numberOfMajor_X_TickMarks))
loglinplot(numpy.array([in_GraphBounds[2], in_GraphBounds[3]]),
numpy.array([in_GraphBounds[0], in_GraphBounds[1]]),
visible=False)
loglinplot(in_DataToPlot[1], in_DataToPlot[0], 'o', markersize=3)
if (min(in_DataToPlot[0]) < in_GraphBounds[0]) or (max(
in_DataToPlot[0]) > in_GraphBounds[1]):
matplotlib.pyplot.ylim(in_GraphBounds[0], in_GraphBounds[1])
if (min(in_DataToPlot[1]) < in_GraphBounds[2]) or (max(
in_DataToPlot[1]) > in_GraphBounds[3]):
matplotlib.pyplot.xlim(in_GraphBounds[2], in_GraphBounds[3])
else:
fig, ax = CommonPlottingCode(in_WidthInPixels, in_HeightInPixels,
in_DataNameX, in_DataNameY,
useOffsetIfNeeded, scientificNotationX,
scientificNotationY, 0.0, 0.0, 1.0, 1.0)
ax.plot(numpy.array([in_GraphBounds[0], in_GraphBounds[1]]),
numpy.array([in_GraphBounds[2], in_GraphBounds[3]
])) # first ax.plot() is only with extents
newLeft, newBottom, newRight, newTop, numberOfMajor_X_TickMarks = YieldNewExtentsAndNumberOfMajor_X_TickMarks(
fig, ax, in_WidthInPixels, in_HeightInPixels, scientificNotationY
or useOffsetIfNeeded)
fig, ax = CommonPlottingCode(in_WidthInPixels, in_HeightInPixels,
in_DataNameX, in_DataNameY,
useOffsetIfNeeded, scientificNotationX,
scientificNotationY, newLeft, newBottom,
newRight, newTop)
if in_LogY == 'LOG' and in_LogX == 'LOG':
loglinplot = ax.loglog
elif in_LogY == 'LIN' and in_LogX == 'LOG':
loglinplot = ax.semilogx
elif in_LogY == 'LOG' and in_LogX == 'LIN':
loglinplot = ax.semilogy
else:
loglinplot = ax.plot
if len(ax.get_xticklabels()) > numberOfMajor_X_TickMarks:
ax.xaxis.set_major_locator(
matplotlib.ticker.MaxNLocator(numberOfMajor_X_TickMarks))
loglinplot(numpy.array([in_GraphBounds[0], in_GraphBounds[1]]),
numpy.array([in_GraphBounds[2], in_GraphBounds[3]]),
visible=False)
loglinplot(in_DataToPlot[0], in_DataToPlot[1], 'o', markersize=3)
if (min(in_DataToPlot[0]) <= in_GraphBounds[0]) or (max(
in_DataToPlot[0]) >= in_GraphBounds[1]):
matplotlib.pyplot.xlim(in_GraphBounds[0], in_GraphBounds[1])
if (min(in_DataToPlot[1]) <= in_GraphBounds[2]) or (max(
in_DataToPlot[1]) >= in_GraphBounds[3]):
matplotlib.pyplot.ylim(in_GraphBounds[2], in_GraphBounds[3])
if in_Equation:
booleanMask = yRange > matplotlib.pyplot.ylim()[0]
booleanMask &= (yRange < matplotlib.pyplot.ylim()[1])
booleanMask &= (xRange > matplotlib.pyplot.xlim()[0])
booleanMask &= (xRange < matplotlib.pyplot.xlim()[1])
loglinplot(xRange[booleanMask], yRange[booleanMask],
'k') # model on top of data points
if inConfidenceIntervalsFlag:
# now calculate confidence intervals for new test x-series
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(in_DataToPlot[0]) # mean of x
n = in_Equation.nobs # number of samples in origional fit
if len(in_Equation.dataCache.allDataCacheDictionary['Weights']
):
t_value = scipy.stats.t.ppf(
0.975, in_Equation.df_e_weighted
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(in_Equation.sumOfSquaredErrors_weighted /
in_Equation.df_e_weighted) *
(1.0 / n +
(numpy.power((xRange - mean_x), 2) /
((numpy.sum(numpy.power(in_DataToPlot[0], 2))) - n *
(numpy.power(mean_x, 2))))))
else:
t_value = scipy.stats.t.ppf(
0.975, in_Equation.df_e
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(in_Equation.sumOfSquaredErrors / in_Equation.df_e) *
(1.0 / n +
(numpy.power((xRange - mean_x), 2) /
((numpy.sum(numpy.power(in_DataToPlot[0], 2))) - n *
(numpy.power(mean_x, 2))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yRange + abs(confs)
lower = yRange - abs(confs)
booleanMask &= (numpy.array(yRange) < 1.0E290)
booleanMask &= (upper < 1.0E290)
booleanMask &= (lower < 1.0E290)
# color scheme improves visibility on black background lines or points
loglinplot(xRange[booleanMask],
lower[booleanMask],
linestyle='solid',
color='white')
loglinplot(xRange[booleanMask],
upper[booleanMask],
linestyle='solid',
color='white')
loglinplot(xRange[booleanMask],
lower[booleanMask],
linestyle='dashed',
color='blue')
loglinplot(xRange[booleanMask],
upper[booleanMask],
linestyle='dashed',
color='blue')
fig.savefig(in_FileNameAndPath[:-3] + 'png', format='png')
if not inPNGOnlyFlag:
fig.savefig(in_FileNameAndPath[:-3] + 'svg', format='svg')
plt.close()
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(
equation.exampleData, equation, False)
equation.Solve()
# raw data for scatterplot
x_data = equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = equation.dataCache.allDataCacheDictionary['IndependentData'][1]
z_data = equation.dataCache.allDataCacheDictionary['DependentData']
# create X, Y, Z mesh grid for surface plot
print "Creating mesh grid data"
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array(
[X, Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
Z = equation.CalculateModelPredictions(
equation.solvedCoefficients, equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
# matplotlib specific code for the plots
fig = plt.figure(figsize=(float(graphWidth) / 100.0,
float(graphHeight) / 100.0),
dpi=100)
ax = fig.gca(projection='3d')
print "Creating matplotlib graph objects"
# create a surface plot using the X, Y, Z mesh data created above
def serialWorker(inputQueue, outputQueue):
for func, args in iter(inputQueue.get, 'STOP'): # iter() has different behaviors depending on number of parameters
outputQueue.put(func(*args))
inputQueue.task_done()
if inputQueue.unfinished_tasks == 0:
break
if __name__ == "__main__":
os.nice(10) ####################### I use this during development
global globalDataCache
globalDataCache = pyeq2.dataCache()
global globalReducedDataCache
globalReducedDataCache = {}
global globalRawData
globalRawData = '''
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
8.442 4.744
9.769 7.068
9.861 7.104
'''
# this example yields a single item to inspect after completion
bestResult = []
# Standard lowest sum-of-squared errors in this example, see IModel.fittingTargetDictionary
fittingTargetText = 'SSQABS'
if fittingTargetText == 'ODR':
raise Exception('ODR cannot use multiple fitting algorithms')
# we are using the same data set repeatedly, so create a cache external to the equations
externalCache = pyeq2.dataCache()
reducedDataCache = {}
#####################################################
# this value is used to make the example run faster #
#####################################################
smoothnessControl = 3
##########################
# fit named equations here
print
print 'Fitting named equations that use non-linear solvers for a fitting target of', fittingTargetText
for submodule in inspect.getmembers(pyeq2.Models_2D):
if inspect.ismodule(submodule[1]):
for equationClass in inspect.getmembers(submodule[1]):
if inspect.isclass(equationClass[1]):
1.408 2.547 1.775
1.279 2.554 1.927
1.016 2.564 2.243
0.742 2.568 2.581
0.607 2.571 2.753
"""
# this example yields a sorted output list to inspect after completion
resultList = []
# Standard lowest sum-of-squared errors in this example, see IModel.fittingTargetDictionary
fittingTargetText = "SSQABS"
# we are using the same data set repeatedly, so create a cache external to the equations
externalCache = pyeq2.dataCache()
reducedDataCache = {}
#####################################################
# this value is used to make the example run faster #
#####################################################
smoothnessControl = 3
##########################
# fit named equations here
for submodule in inspect.getmembers(pyeq2.Models_3D):
if inspect.ismodule(submodule[1]):
for equationClass in inspect.getmembers(submodule[1]):
if inspect.isclass(equationClass[1]):
def __init__(self, inFittingTarget = 'SSQABS', inExtendedVersionName = 'Default'):
if inExtendedVersionName == '':
inExtendedVersionName = 'Default'
if inFittingTarget not in list(self.fittingTargetDictionary.keys()):
raise Exception(str(inFittingTarget) + ' is not in the IModel class fitting target dictionary.')
self.fittingTarget = inFittingTarget
inExtendedVersionName = inExtendedVersionName.replace(' ', '')
if inExtendedVersionName not in pyeq2.ExtendedVersionHandlers.extendedVersionHandlerNameList:
raise Exception(inExtendedVersionName + ' is not in the list of extended version handler names.')
allowedExtendedVersion = True
if (-1 != inExtendedVersionName.find('Offset')) and (self.autoGenerateOffsetForm == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Reciprocal')) and (self.autoGenerateReciprocalForm == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Inverse')) and (self.autoGenerateInverseForms == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Growth')) and (self.autoGenerateGrowthAndDecayForms == False):
allowedExtendedVersion = False
if (-1 != inExtendedVersionName.find('Decay')) and (self.autoGenerateGrowthAndDecayForms == False):
allowedExtendedVersion = False
if allowedExtendedVersion == False:
raise Exception('This equation does not allow an extended version named "' + inExtendedVersionName + '".')
self.extendedVersionHandler = eval('pyeq2.ExtendedVersionHandlers.ExtendedVersionHandler_' + inExtendedVersionName + '.ExtendedVersionHandler_' + inExtendedVersionName + '()')
self.dataCache = pyeq2.dataCache()
self.upperCoefficientBounds = []
self.lowerCoefficientBounds = []
self.estimatedCoefficients = []
self.fixedCoefficients = []
self.solvedCoefficients = []
self.polyfunctional2DFlags = []
self.polyfunctional3DFlags = []
self.xPolynomialOrder = None
self.yPolynomialOrder = None
self.rationalNumeratorFlags = []
self.rationalDenominatorFlags = []
self.deEstimatedCoefficients = []
try:
if self._dimensionality == 2:
self.exampleData = '''
X Y
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
9.769 7.068
9.861 7.104
'''
else:
self.exampleData = '''
X Y Z
3.017 2.175 0.320
2.822 2.624 0.629
2.632 2.839 0.950
2.287 3.030 1.574
2.207 3.057 1.725
2.048 3.098 2.035
1.963 3.115 2.204
1.784 3.144 2.570
1.712 3.153 2.721
2.972 2.106 0.313
2.719 2.542 0.643
2.495 2.721 0.956
2.070 2.878 1.597
1.969 2.899 1.758
1.768 2.929 2.088
1.677 2.939 2.240
1.479 2.957 2.583
1.387 2.963 2.744
2.843 1.984 0.315
2.485 2.320 0.639
2.163 2.444 0.954
1.687 2.525 1.459
1.408 2.547 1.775
1.279 2.554 1.927
1.016 2.564 2.243
0.742 2.568 2.581
0.607 2.571 2.753
'''
except:
pass
def draw(self, equation):
if not self.parent.y_data:
self.parent.y_data = equation.dataCache.allDataCacheDictionary[
'DependentData']
if not self.parent.x_data:
self.parent.x_data = equation.dataCache.allDataCacheDictionary[
'IndependentData'][0]
# now create data for the fitted equation plot
xModel = numpy.linspace(min(self.parent.x_data),
max(self.parent.x_data))
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([xModel, xModel])
equation.dataCache.FindOrCreateAllDataCache(equation)
yModel = equation.CalculateModelPredictions(
equation.solvedCoefficients,
equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
# first the raw data as a scatter plot
self.axes.plot(self.parent.x_data, self.parent.y_data, 'D')
# now the model as a line plot
self.axes.plot(xModel, yModel)
# now calculate confidence intervals
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(self.parent.x_data) # mean of x
n = equation.nobs # number of samples in the origional fit
t_value = scipy.stats.t.ppf(
0.975, equation.df_e
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(equation.sumOfSquaredErrors / equation.df_e) *
(1.0 / n + (numpy.power((xModel - mean_x), 2.0) /
((numpy.sum(numpy.power(self.parent.x_data, 2))) - n *
(numpy.power(mean_x, 2.0))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yModel + abs(confs)
lower = yModel - abs(confs)
# mask off any numbers outside the existing plot limits
booleanMask = yModel > matplotlib.pyplot.ylim()[0]
booleanMask &= (yModel < matplotlib.pyplot.ylim()[1])
# color scheme improves visibility on black background lines or points
self.axes.plot(xModel[booleanMask],
lower[booleanMask],
linestyle='solid',
color='white')
self.axes.plot(xModel[booleanMask],
upper[booleanMask],
linestyle='solid',
color='white')
self.axes.plot(xModel[booleanMask],
lower[booleanMask],
linestyle='dashed',
color='blue')
self.axes.plot(xModel[booleanMask],
upper[booleanMask],
linestyle='dashed',
color='blue')
self.axes.set_title('Model With 95% Confidence Limits') # add a title
self.axes.set_xlabel('X Data') # X axis data label
self.axes.set_ylabel('Y Data') # Y axis data label
def SurfaceAndContourPlots(in_filePathSurface, in_filePathContour, in_equation,
in_surfaceTitle, in_contourTitle, in_xAxisLabel,
in_yAxisLabel, in_zAxisLabel):
# raw data
x_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][1]
z_data = in_equation.dataCache.allDataCacheDictionary['DependentData']
from mpl_toolkits.mplot3d import Axes3D # 3D apecific
from matplotlib import cm # to colormap from blue to red
fig = plt.figure()
ax = fig.gca(projection='3d')
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = in_equation.dataCache
in_equation.dataCache = pyeq2.dataCache()
in_equation.dataCache.allDataCacheDictionary[
'IndependentData'] = numpy.array([X, Y])
in_equation.dataCache.FindOrCreateAllDataCache(in_equation)
Z = in_equation.CalculateModelPredictions(
in_equation.solvedCoefficients,
in_equation.dataCache.allDataCacheDictionary)
in_equation.dataCache = tempcache
ax.plot_surface(X,
Y,
Z,
rstride=1,
cstride=1,
cmap=cm.coolwarm,
linewidth=1,
antialiased=True)
ax.scatter(x_data, y_data, z_data)
ax.set_title(in_surfaceTitle) # add a title for surface plot
ax.set_xlabel(in_xAxisLabel) # X axis data label
ax.set_ylabel(in_yAxisLabel) # Y axis data label
ax.set_zlabel(in_zAxisLabel) # Y axis data label
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePathSurface) # create PNG file
plt.close('all')
# contour plot here
fig = plt.figure(figsize=(5, 4))
ax = fig.add_subplot(1, 1, 1)
ax.plot(
x_data, y_data, 'o', color='0.8', markersize=4
) # draw these first so contour lines overlay. Color=number is grayscale
ax.set_title(in_contourTitle) # add a title for contour plot
ax.set_xlabel(in_xAxisLabel) # X data label
ax.set_ylabel(in_yAxisLabel) # Y data label
numberOfContourLines = 16
CS = plt.contour(X, Y, Z, numberOfContourLines, colors='k')
plt.clabel(CS, inline=1, fontsize=10) # labels for contours
plt.tight_layout() # prevents cropping axis labels
fig.savefig(in_filePathContour) # create PNG file
plt.close('all')
def ContourPlot(in_DataObject, in_FileNameAndPath):
gridResolution = (in_DataObject.graphWidth + in_DataObject.graphHeight) / 20
gxmin = in_DataObject.gxmin
gxmax = in_DataObject.gxmax
gymin = in_DataObject.gymin
gymax = in_DataObject.gymax
if in_DataObject.equation.independentData1CannotContainNegativeFlag and gxmin < 0.0:
gxmin = 0.0
if in_DataObject.equation.independentData1CannotContainZeroFlag and gxmin == 0.0:
gxmin = 1.0e-300
if in_DataObject.equation.independentData1CannotContainPositiveFlag and gxmax > 0.0:
gxmax = 0.0
if in_DataObject.equation.independentData1CannotContainZeroFlag and gxmax == 0.0:
gxmax = 1.0e-300
if in_DataObject.equation.independentData2CannotContainNegativeFlag and gymin < 0.0:
gymin = 0.0
if in_DataObject.equation.independentData2CannotContainZeroFlag and gymin == 0.0:
gymin = 1.0e-300
if in_DataObject.equation.independentData2CannotContainPositiveFlag and gymax > 0.0:
gymax = 0.0
if in_DataObject.equation.independentData2CannotContainZeroFlag and gymax == 0.0:
gymax = 1.0e-300
deltax = (gxmax - gxmin) / float(gridResolution)
deltay = (gymax - gymin) / float(gridResolution)
xRange = numpy.arange(gxmin, gxmax + deltax, deltax)
yRange = numpy.arange(gymin, gymax + deltay / 2.0, deltay)
minZ = min(in_DataObject.DependentDataArray)
maxZ = max(in_DataObject.DependentDataArray)
X, Y = numpy.meshgrid(xRange, yRange)
boundingRectangle = matplotlib.patches.Rectangle(
[gxmin, gymin], gxmax - gxmin, gymax - gymin, facecolor=(0.975, 0.975, 0.975), edgecolor=(0.9, 0.9, 0.9)
)
Z = []
tempDataCache = in_DataObject.equation.dataCache
for i in range(len(X)):
in_DataObject.equation.dataCache = pyeq2.dataCache()
in_DataObject.equation.dataCache.allDataCacheDictionary["IndependentData"] = numpy.array([X[i], Y[i]])
in_DataObject.equation.dataCache.FindOrCreateAllDataCache(in_DataObject.equation)
Z.append(
in_DataObject.equation.CalculateModelPredictions(
in_DataObject.equation.solvedCoefficients, in_DataObject.equation.dataCache.allDataCacheDictionary
)
)
in_DataObject.equation.dataCache = tempDataCache
Z = numpy.array(Z)
Z = numpy.clip(Z, minZ, maxZ)
tempData = [
in_DataObject.IndependentDataArray[0],
in_DataObject.IndependentDataArray[1],
in_DataObject.DependentDataArray,
]
ContourPlot_NoDataObject(
X,
Y,
Z,
tempData,
in_FileNameAndPath,
in_DataObject.IndependentDataName1,
in_DataObject.IndependentDataName2,
in_DataObject.graphWidth,
in_DataObject.graphHeight,
"UseOffset_ON",
"ScientificNotation_X_AUTO",
"ScientificNotation_Y_AUTO",
in_DataObject.pngOnlyFlag,
boundingRectangle,
)
plt.close()
def ScatterPlotWithOptionalModel_NoDataObject(
in_DataToPlot,
in_FileNameAndPath,
in_DataNameX,
in_DataNameY,
in_WidthInPixels,
in_HeightInPixels,
in_Equation,
in_UseOffsetIfNeeded,
in_ReverseXY,
in_X_UseScientificNotationIfNeeded,
in_Y_UseScientificNotationIfNeeded,
in_GraphBounds,
in_LogY,
in_LogX,
inPNGOnlyFlag,
inConfidenceIntervalsFlag,
):
# decode ends of strings ('XYZ_ON', 'XYZ_OFF', 'XYZ_AUTO', etc.) to boolean values
scientificNotationX = DetermineScientificNotationFromString(in_DataToPlot[0], in_X_UseScientificNotationIfNeeded)
scientificNotationY = DetermineScientificNotationFromString(in_DataToPlot[1], in_Y_UseScientificNotationIfNeeded)
useOffsetIfNeeded = DetermineOnOrOffFromString(in_UseOffsetIfNeeded)
reverseXY = DetermineOnOrOffFromString(in_ReverseXY)
if in_Equation: # make model data for plotting, clipping at boundaries
lowerXbound = in_GraphBounds[0]
upperXbound = in_GraphBounds[1]
if in_Equation.independentData1CannotContainNegativeFlag and lowerXbound < 0.0:
lowerXbound = 0.0
if in_Equation.independentData1CannotContainZeroFlag and lowerXbound == 0.0:
lowerXbound = 1.0e-300
if in_Equation.independentData1CannotContainPositiveFlag and upperXbound > 0.0:
upperXbound = 0.0
if in_Equation.independentData1CannotContainZeroFlag and upperXbound == 0.0:
upperXbound = 1.0e-300
xRange = numpy.arange(
lowerXbound, upperXbound, (upperXbound - lowerXbound) / (20.0 * float(in_WidthInPixels + in_HeightInPixels))
) # make this 'reverse-xy-independent'
tempDataCache = in_Equation.dataCache
in_Equation.dataCache = pyeq2.dataCache()
in_Equation.dataCache.allDataCacheDictionary["IndependentData"] = numpy.array([xRange, xRange])
in_Equation.dataCache.FindOrCreateAllDataCache(in_Equation)
yRange = in_Equation.CalculateModelPredictions(
in_Equation.solvedCoefficients, in_Equation.dataCache.allDataCacheDictionary
)
in_Equation.dataCache = tempDataCache
if reverseXY:
fig, ax = CommonPlottingCode(
in_WidthInPixels,
in_HeightInPixels,
in_DataNameX,
in_DataNameY,
useOffsetIfNeeded,
scientificNotationY,
scientificNotationX,
0.0,
0.0,
1.0,
1.0,
)
ax.plot(
numpy.array([in_GraphBounds[2], in_GraphBounds[3]]), numpy.array([in_GraphBounds[0], in_GraphBounds[1]])
) # first ax.plot() is only with extents
newLeft, newBottom, newRight, newTop, numberOfMajor_X_TickMarks = YieldNewExtentsAndNumberOfMajor_X_TickMarks(
fig, ax, in_WidthInPixels, in_HeightInPixels, scientificNotationY or useOffsetIfNeeded
)
fig, ax = CommonPlottingCode(
in_WidthInPixels,
in_HeightInPixels,
in_DataNameX,
in_DataNameY,
useOffsetIfNeeded,
scientificNotationY,
scientificNotationX,
newLeft,
newBottom,
newRight,
newTop,
)
if in_LogY == "LOG" and in_LogX == "LOG":
loglinplot = ax.loglog
elif in_LogY == "LIN" and in_LogX == "LOG":
loglinplot = ax.semilogx
elif in_LogY == "LOG" and in_LogX == "LIN":
loglinplot = ax.semilogy
else:
loglinplot = ax.plot
if len(ax.get_xticklabels()) > numberOfMajor_X_TickMarks:
ax.xaxis.set_major_locator(matplotlib.ticker.MaxNLocator(numberOfMajor_X_TickMarks))
loglinplot(
numpy.array([in_GraphBounds[2], in_GraphBounds[3]]),
numpy.array([in_GraphBounds[0], in_GraphBounds[1]]),
visible=False,
)
loglinplot(in_DataToPlot[1], in_DataToPlot[0], "o", markersize=3)
if (min(in_DataToPlot[0]) < in_GraphBounds[0]) or (max(in_DataToPlot[0]) > in_GraphBounds[1]):
matplotlib.pyplot.ylim(in_GraphBounds[0], in_GraphBounds[1])
if (min(in_DataToPlot[1]) < in_GraphBounds[2]) or (max(in_DataToPlot[1]) > in_GraphBounds[3]):
matplotlib.pyplot.xlim(in_GraphBounds[2], in_GraphBounds[3])
else:
fig, ax = CommonPlottingCode(
in_WidthInPixels,
in_HeightInPixels,
in_DataNameX,
in_DataNameY,
useOffsetIfNeeded,
scientificNotationX,
scientificNotationY,
0.0,
0.0,
1.0,
1.0,
)
ax.plot(
numpy.array([in_GraphBounds[0], in_GraphBounds[1]]), numpy.array([in_GraphBounds[2], in_GraphBounds[3]])
) # first ax.plot() is only with extents
newLeft, newBottom, newRight, newTop, numberOfMajor_X_TickMarks = YieldNewExtentsAndNumberOfMajor_X_TickMarks(
fig, ax, in_WidthInPixels, in_HeightInPixels, scientificNotationY or useOffsetIfNeeded
)
fig, ax = CommonPlottingCode(
in_WidthInPixels,
in_HeightInPixels,
in_DataNameX,
in_DataNameY,
useOffsetIfNeeded,
scientificNotationX,
scientificNotationY,
newLeft,
newBottom,
newRight,
newTop,
)
if in_LogY == "LOG" and in_LogX == "LOG":
loglinplot = ax.loglog
elif in_LogY == "LIN" and in_LogX == "LOG":
loglinplot = ax.semilogx
elif in_LogY == "LOG" and in_LogX == "LIN":
loglinplot = ax.semilogy
else:
loglinplot = ax.plot
if len(ax.get_xticklabels()) > numberOfMajor_X_TickMarks:
ax.xaxis.set_major_locator(matplotlib.ticker.MaxNLocator(numberOfMajor_X_TickMarks))
loglinplot(
numpy.array([in_GraphBounds[0], in_GraphBounds[1]]),
numpy.array([in_GraphBounds[2], in_GraphBounds[3]]),
visible=False,
)
loglinplot(in_DataToPlot[0], in_DataToPlot[1], "o", markersize=3)
if (min(in_DataToPlot[0]) <= in_GraphBounds[0]) or (max(in_DataToPlot[0]) >= in_GraphBounds[1]):
matplotlib.pyplot.xlim(in_GraphBounds[0], in_GraphBounds[1])
if (min(in_DataToPlot[1]) <= in_GraphBounds[2]) or (max(in_DataToPlot[1]) >= in_GraphBounds[3]):
matplotlib.pyplot.ylim(in_GraphBounds[2], in_GraphBounds[3])
if in_Equation:
booleanMask = yRange > matplotlib.pyplot.ylim()[0]
booleanMask &= yRange < matplotlib.pyplot.ylim()[1]
booleanMask &= xRange > matplotlib.pyplot.xlim()[0]
booleanMask &= xRange < matplotlib.pyplot.xlim()[1]
loglinplot(xRange[booleanMask], yRange[booleanMask], "k") # model on top of data points
if inConfidenceIntervalsFlag:
# now calculate confidence intervals for new test x-series
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(in_DataToPlot[0]) # mean of x
n = in_Equation.nobs # number of samples in origional fit
if len(in_Equation.dataCache.allDataCacheDictionary["Weights"]):
t_value = scipy.stats.t.ppf(
0.975, in_Equation.df_e_weighted
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(in_Equation.sumOfSquaredErrors_weighted / in_Equation.df_e_weighted)
* (
1.0 / n
+ (
numpy.power((xRange - mean_x), 2)
/ ((numpy.sum(numpy.power(in_DataToPlot[0], 2))) - n * (numpy.power(mean_x, 2)))
)
)
)
else:
t_value = scipy.stats.t.ppf(
0.975, in_Equation.df_e
) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt(
(in_Equation.sumOfSquaredErrors / in_Equation.df_e)
* (
1.0 / n
+ (
numpy.power((xRange - mean_x), 2)
/ ((numpy.sum(numpy.power(in_DataToPlot[0], 2))) - n * (numpy.power(mean_x, 2)))
)
)
)
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yRange + abs(confs)
lower = yRange - abs(confs)
booleanMask &= numpy.array(yRange) < 1.0e290
booleanMask &= upper < 1.0e290
booleanMask &= lower < 1.0e290
# color scheme improves visibility on black background lines or points
loglinplot(xRange[booleanMask], lower[booleanMask], linestyle="solid", color="white")
loglinplot(xRange[booleanMask], upper[booleanMask], linestyle="solid", color="white")
loglinplot(xRange[booleanMask], lower[booleanMask], linestyle="dashed", color="blue")
loglinplot(xRange[booleanMask], upper[booleanMask], linestyle="dashed", color="blue")
fig.savefig(in_FileNameAndPath[:-3] + "png", format="png")
if not inPNGOnlyFlag:
fig.savefig(in_FileNameAndPath[:-3] + "svg", format="svg")
plt.close()
def ParallelFittingFunction(rawData, fittingTargetText, smoothnessControl, modulus, modulusRemainder):
processID = str(os.getpid())
# this function yields a single item to inspect after completion
bestResult = []
# we are using the same data set repeatedly, so create a cache external to the equations
externalCache = pyeq2.dataCache()
reducedDataCache = {}
equationCountForModulus = 0
##########################
# add named equations here
print()
print('Process ID', processID, 'fitting named equations:')
for submodule in inspect.getmembers(pyeq2.Models_2D):
if inspect.ismodule(submodule[1]):
for equationClass in inspect.getmembers(submodule[1]):
if inspect.isclass(equationClass[1]):
# special classes
if equationClass[1].splineFlag or \
equationClass[1].userSelectablePolynomialFlag or \
equationClass[1].userCustomizablePolynomialFlag or \
equationClass[1].userSelectablePolyfunctionalFlag or \
equationClass[1].userSelectableRationalFlag or \
equationClass[1].userDefinedFunctionFlag:
continue
for extendedVersion in ['Default', 'Offset']:
if (extendedVersion == 'Offset') and (equationClass[1].autoGenerateOffsetForm == False):
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = equationClass[1](fittingTargetText, extendedVersion)
if len(equationInstance.GetCoefficientDesignators()) > smoothnessControl:
continue
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(equationInstance)
if reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, True, processID)
if result:
bestResult = result
if not reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
reducedDataCache[equationInstance.numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
##########################
# fit polyfunctionals here
()
print('Process ID', processID, 'fitting polyfunctionals:')
equationCount = 0
maxPolyfunctionalCoefficients = 4 # this value was chosen to make this example more convenient
polyfunctionalEquationList = pyeq2.PolyFunctions.GenerateListForPolyfunctionals_2D()
functionIndexList = range(len(polyfunctionalEquationList)) # make a list of function indices to permute
for coeffCount in range(1, maxPolyfunctionalCoefficients+1):
functionCombinations = UniqueCombinations(functionIndexList, coeffCount)
for functionCombination in functionCombinations:
if len(functionCombination) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Polyfunctional.UserSelectablePolyfunctional(fittingTargetText, 'Default', functionCombination, polyfunctionalEquationList)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(equationInstance)
if reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
reducedDataCache[equationInstance.numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
equationCount += 1
if (equationCount % 50) == 0:
print(' Process ID', processID, 'fitted', equationCount, 'equations...')
######################
# fit user-selectable polynomials here
print()
print('Process ID', processID, 'fitting user-selectable polynomials:')
maxPolynomialOrderX = 5 # this value was chosen to make this example more convenient
for polynomialOrderX in range(maxPolynomialOrderX+1):
if (polynomialOrderX + 1) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Polynomial.UserSelectablePolynomial(fittingTargetText, 'Default', polynomialOrderX)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(equationInstance)
if reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
reducedDataCache[equationInstance.numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
######################
# fit user-selectable rationals here
print
print('Process ID', processID, 'fitting user-selectable rationals:')
equationCount = 0
maxCoeffs = smoothnessControl # arbitrary choice of maximum total coefficients for this example
functionList = pyeq2.PolyFunctions.GenerateListForRationals_2D()
functionIndexList = range(len(functionList)) # make a list of function indices
for numeratorCoeffCount in range(1, maxCoeffs):
numeratorComboList = UniqueCombinations(functionIndexList, numeratorCoeffCount)
for numeratorCombo in numeratorComboList:
for denominatorCoeffCount in range(1, maxCoeffs):
denominatorComboList = UniqueCombinations2(functionIndexList, denominatorCoeffCount)
for denominatorCombo in denominatorComboList:
for extendedVersion in ['Default', 'Offset']:
extraCoeffs = 0
if extendedVersion == 'Offset':
extraCoeffs = 1
if (len(numeratorCombo) + len(denominatorCombo) + extraCoeffs) > smoothnessControl:
continue
if equationCountForModulus % modulus != modulusRemainder:
equationCountForModulus += 1
continue
equationCountForModulus += 1
equationInstance = pyeq2.Models_2D.Rational.UserSelectableRational(fittingTargetText, extendedVersion, numeratorCombo, denominatorCombo, functionList)
equationInstance.dataCache = externalCache # re-use the external cache
if equationInstance.dataCache.allDataCacheDictionary == {}:
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(rawData, equationInstance, False)
equationInstance.dataCache.CalculateNumberOfReducedDataPoints(equationInstance)
if reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
equationInstance.dataCache.reducedDataCacheDictionary = reducedDataCache[equationInstance.numberOfReducedDataPoints]
else:
equationInstance.dataCache.reducedDataCacheDictionary = {}
result = SetParametersAndFit(equationInstance, bestResult, False)
if result:
bestResult = result
if not reducedDataCache.has_key(equationInstance.numberOfReducedDataPoints):
reducedDataCache[equationInstance.numberOfReducedDataPoints] = equationInstance.dataCache.reducedDataCacheDictionary
equationCount += 1
if (equationCount % 5) == 0:
print(' ', 'Process ID', processID + ',', equationCount, 'rationals, current flags:', equationInstance.rationalNumeratorFlags, equationInstance.rationalDenominatorFlags,)
if extendedVersion == 'Offset':
print('with offset')
else:
print()
print('Process ID', processID, 'has completed')
return bestResult
def SaveModelScatterConfidence(in_fileName, in_equation, in_Ymax, in_Ymin):
# raw data
x_data = in_equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = in_equation.dataCache.allDataCacheDictionary['DependentData']
# now create data for the fitted in_equation plot
xModel = numpy.linspace(min(x_data), max(x_data))
tempcache = in_equation.dataCache
in_equation.dataCache = pyeq2.dataCache()
in_equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([xModel, xModel])
in_equation.dataCache.FindOrCreateAllDataCache(in_equation)
yModel = in_equation.CalculateModelPredictions(in_equation.solvedCoefficients, in_equation.dataCache.allDataCacheDictionary)
in_equation.dataCache = tempcache
# now use matplotlib to create the PNG file
if -1 != in_fileName.find('_large'): # large animation frame
fig = plt.figure(figsize=(float(largeAnimationGraphWidth ) / 100.0, float(largeAnimationGraphHeight ) / 100.0), dpi=100)
else: # small animation frame
fig = plt.figure(figsize=(float(smallAnimationGraphWidth ) / 100.0, float(smallAnimationGraphHeight ) / 100.0), dpi=100)
ax = fig.add_subplot(1,1,1)
# first the raw data as a scatter plot
if -1 != in_fileName.find('_small'): # small animation frame
ax.plot(x_data, y_data, 'D', markersize = 2)
else:
ax.plot(x_data, y_data, 'D')
# now the model as a line plot
ax.plot(xModel, yModel, label = 'Fitted Equation')
# now calculate confidence intervals
# http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_nlin_sect026.htm
# http://www.staff.ncl.ac.uk/tom.holderness/software/pythonlinearfit
mean_x = numpy.mean(x_data) # mean value of x
n = len(x_data) # number of samples
t_value = scipy.stats.t.ppf(0.975, in_equation.df_e) # (1.0 - (a/2)) is used for two-sided t-test critical value, here a = 0.05
confs = t_value * numpy.sqrt((in_equation.sumOfSquaredErrors/in_equation.df_e)*(1.0/n + (numpy.power((xModel-mean_x),2.0)/
((numpy.sum(numpy.power(x_data,2)))-n*(numpy.power(mean_x,2.0))))))
# get lower and upper confidence limits based on predicted y and confidence intervals
upper = yModel + abs(confs)
lower = yModel - abs(confs)
# mask off any numbers outside the existing plot limits
booleanMask = yModel > matplotlib.pyplot.ylim()[0]
booleanMask &= (yModel < matplotlib.pyplot.ylim()[1])
# color scheme improves visibility on black background lines or points
ax.plot(xModel[booleanMask], lower[booleanMask], linestyle='solid', color='white')
ax.plot(xModel[booleanMask], upper[booleanMask], linestyle='solid', color='white')
ax.plot(xModel[booleanMask], lower[booleanMask], linestyle='dashed', color='blue')
ax.plot(xModel[booleanMask], upper[booleanMask], linestyle='dashed', color='blue', label = '95% Confidence Intervals')
if -1 != in_fileName.find('_small'): # small animation frame
ax.set_ylabel("Y Data", size='small') # Y axis data label
ax.set_xlabel("X Data", size='small') # X axis data label
else:
ax.set_ylabel("Y Data") # Y axis data label
ax.set_xlabel("X Data") # X axis data label
if -1 != in_fileName.find('_small'): # small animation frame
for xlabel_i in ax.get_xticklabels():
xlabel_i.set_fontsize(xlabel_i.get_fontsize() * 0.5)
for ylabel_i in ax.get_yticklabels():
ylabel_i.set_fontsize(ylabel_i.get_fontsize() * 0.5)
plt.ylim(in_Ymin, in_Ymax)
# legends on large images
# http://matplotlib.org/api/pyplot_api.html#matplotlib.pyplot.legend
if -1 != in_fileName.find('large'): # large animation frame
ax.legend(loc=2, fontsize='small')
fig.tight_layout()
fig.savefig(in_fileName) # create PNG file
plt.close('all') # clear pyplot else memory use becomes large
def GenerateListOfOutputReports(self):
if self.debug: print "**** FF Results GenerateListOfOutputReports 1"
self.textReports = []
self.graphReports = []
externalDataCache = pyeq2.dataCache(
) # reuse this to speed up some caching
for i in range(self.numberOfEquationsToDisplay):
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.1: i=", str(
i)
listItem = self.functionFinderResultsList[i + self.rank - 1]
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.2: i=", str(
i)
reportDataObject = copy.deepcopy(self.dataObject)
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.3: i=", str(
i)
# find the equation instance for the incoming dimensionality, equation family name and equation name - 404 if not found
reportDataObject.equation = eval(listItem[1] + "." + listItem[2] +
"('SSQABS', '" + listItem[3] +
"')")
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.4: i=", str(
i
), 'equation name =', reportDataObject.equation.GetDisplayName(
)
if externalDataCache.allDataCacheDictionary == {}: # This should only run for the first equation
temp = reportDataObject.textDataEditor
# comma conversions
if reportDataObject.commaConversion == "D": # decimal separator
temp = temp.replace(",", ".")
elif reportDataObject.commaConversion == "I": # as if they don't exist
temp = temp.replace(",", "")
else:
temp = temp.replace(
",", " ") # default to the original default conversion
# replace these characters with spaces for use by float()
temp = temp.replace("$", " ")
temp = temp.replace("%", " ")
temp = temp.replace("(", " ")
temp = temp.replace(")", " ")
temp = temp.replace("{", " ")
temp = temp.replace("}", " ")
temp = temp.replace('\r\n', '\n')
temp = temp.replace('\r', '\n')
# replace HTML spaces and tabs with spaces
temp = temp.replace(" ", " ")
temp = temp.replace("	", " ")
temp = temp.replace("	", " ")
temp = temp.replace(" ", " ")
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(
temp, reportDataObject.equation, False)
externalDataCache = reportDataObject.equation.dataCache
reportDataObject.equation.polyfunctional2DFlags = listItem[4]
reportDataObject.equation.polyfunctional3DFlags = listItem[5]
reportDataObject.equation.xPolynomialOrder = listItem[6]
reportDataObject.equation.yPolynomialOrder = listItem[7]
reportDataObject.equation.rationalNumeratorFlags = listItem[8]
reportDataObject.equation.rationalDenominatorFlags = listItem[9]
reportDataObject.equation.fittingTarget = listItem[10]
reportDataObject.equation.solvedCoefficients = listItem[11]
targetValue = listItem[0]
reportDataObject.equation.dataCache = externalDataCache
reportDataObject.equation.dataCache.FindOrCreateAllDataCache(
reportDataObject.equation)
externalDataCache = reportDataObject.equation.dataCache
# add a bit more extrapolation for the function finder result displays
reportDataObject.Extrapolation_x = 0.05
reportDataObject.Extrapolation_y = 0.05
reportDataObject.Extrapolation_z = 0.05
# needed here for graph boundary calculation
reportDataObject.graphWidth = 280
reportDataObject.graphHeight = 240
# 3D rotation angles
reportDataObject.altimuth3D = 45.0
reportDataObject.azimuth3D = 45.0
reportDataObject.CalculateDataStatistics()
reportDataObject.CalculateErrorStatistics()
reportDataObject.CalculateGraphBoundaries()
reportDataObject.equation.CalculateCoefficientAndFitStatistics()
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.5: i=", str(
i
), 'equation name =', reportDataObject.equation.GetDisplayName(
)
graphs = []
# Different graphs for 2D and 3D
reportDataObject.pngOnlyFlag = True
if reportDataObject.equation.GetDimensionality() == 2:
graph = ReportsAndGraphs.DependentDataVsIndependentData1_ModelPlot(
reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
else:
graph = ReportsAndGraphs.SurfacePlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
graph = ReportsAndGraphs.ContourPlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
graphs.append(graph.websiteFileLocation)
self.graphReports.append(graph)
if reportDataObject.equation.fittingTarget[-3:] != "REL":
graph = ReportsAndGraphs.AbsoluteErrorVsDependentData_ScatterPlot(
reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
else:
graph = ReportsAndGraphs.RelativeErrorVsDependentData_ScatterPlot(
reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
graphs.append(graph.websiteFileLocation)
self.graphReports.append(graph)
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.6: i=", str(
i
), 'equation name =', reportDataObject.equation.GetDisplayName(
)
dataForOneEquation = {}
splitted = listItem[1].split('.')
dataForOneEquation['moduleName'] = splitted[-1]
dataForOneEquation[
'displayName'] = reportDataObject.equation.GetDisplayName()
dataForOneEquation['URLQuotedModuleName'] = urllib.quote(
splitted[-1])
dataForOneEquation['URLQuotedDisplayName'] = urllib.quote(
reportDataObject.equation.GetDisplayName())
dataForOneEquation[
'displayHTML'] = reportDataObject.equation.GetDisplayHTML()
dataForOneEquation['graphWebSiteLocations'] = graphs
dataForOneEquation['rank'] = i + self.rank
dataForOneEquation['dimensionality'] = self.dimensionality
dataForOneEquation[
'fittingTarget'] = reportDataObject.equation.fittingTarget
dataForOneEquation['fittingTargetValue'] = targetValue
if reportDataObject.fittingTarget[
-3:] != "REL": # only non-relative error fits get RMSE displayed
dataForOneEquation['rmseString'] = '<br>RMSE: ' + str(
reportDataObject.equation.rmse) + '<br>'
else:
dataForOneEquation['rmseString'] = '<br>'
self.equationDataForDjangoTemplate.append(dataForOneEquation)
if self.debug:
print "**** FF Results GenerateListOfOutputReports 2.7: i=", str(
i
), 'equation name =', reportDataObject.equation.GetDisplayName(
)
def GenerateListOfOutputReports(self):
if self.debug: print "**** FF Results GenerateListOfOutputReports 1"
self.textReports = []
self.graphReports = []
externalDataCache = pyeq2.dataCache() # reuse this to speed up some caching
for i in range(self.numberOfEquationsToDisplay):
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.1: i=", str(i)
listItem = self.functionFinderResultsList[i + self.rank-1]
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.2: i=", str(i)
reportDataObject = copy.deepcopy(self.dataObject)
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.3: i=", str(i)
# find the equation instance for the incoming dimensionality, equation family name and equation name - 404 if not found
reportDataObject.equation = eval(listItem[1] + "." + listItem[2] + "('SSQABS', '" + listItem[3] + "')")
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.4: i=", str(i), 'equation name =', reportDataObject.equation.GetDisplayName()
if externalDataCache.allDataCacheDictionary == {}: # This should only run for the first equation
temp = reportDataObject.textDataEditor
# comma conversions
if reportDataObject.commaConversion == "D": # decimal separator
temp = temp.replace(",",".")
elif reportDataObject.commaConversion == "I": # as if they don't exist
temp = temp.replace(",","")
else:
temp = temp.replace(","," ") # default to the original default conversion
# replace these characters with spaces for use by float()
temp = temp.replace("$"," ")
temp = temp.replace("%"," ")
temp = temp.replace("("," ")
temp = temp.replace(")"," ")
temp = temp.replace("{"," ")
temp = temp.replace("}"," ")
temp = temp.replace('\r\n','\n')
temp = temp.replace('\r','\n')
# replace HTML spaces and tabs with spaces
temp = temp.replace(" "," ")
temp = temp.replace("	"," ")
temp = temp.replace("	"," ")
temp = temp.replace(" "," ")
pyeq2.dataConvertorService().ConvertAndSortColumnarASCII(temp, reportDataObject.equation, False)
externalDataCache = reportDataObject.equation.dataCache
reportDataObject.equation.polyfunctional2DFlags = listItem[4]
reportDataObject.equation.polyfunctional3DFlags = listItem[5]
reportDataObject.equation.xPolynomialOrder = listItem[6]
reportDataObject.equation.yPolynomialOrder = listItem[7]
reportDataObject.equation.rationalNumeratorFlags = listItem[8]
reportDataObject.equation.rationalDenominatorFlags = listItem[9]
reportDataObject.equation.fittingTarget = listItem[10]
reportDataObject.equation.solvedCoefficients = listItem[11]
targetValue = listItem[0]
reportDataObject.equation.dataCache = externalDataCache
reportDataObject.equation.dataCache.FindOrCreateAllDataCache(reportDataObject.equation)
externalDataCache = reportDataObject.equation.dataCache
# add a bit more extrapolation for the function finder result displays
reportDataObject.Extrapolation_x = 0.05
reportDataObject.Extrapolation_y = 0.05
reportDataObject.Extrapolation_z = 0.05
# needed here for graph boundary calculation
reportDataObject.graphWidth = 280
reportDataObject.graphHeight = 240
# 3D rotation angles
reportDataObject.altimuth3D = 45.0
reportDataObject.azimuth3D = 45.0
reportDataObject.CalculateDataStatistics()
reportDataObject.CalculateErrorStatistics()
reportDataObject.CalculateGraphBoundaries()
reportDataObject.equation.CalculateCoefficientAndFitStatistics()
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.5: i=", str(i), 'equation name =', reportDataObject.equation.GetDisplayName()
graphs = []
# Different graphs for 2D and 3D
reportDataObject.pngOnlyFlag = True
if reportDataObject.equation.GetDimensionality() == 2:
graph = ReportsAndGraphs.DependentDataVsIndependentData1_ModelPlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
else:
graph = ReportsAndGraphs.SurfacePlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
graph = ReportsAndGraphs.ContourPlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
graphs.append(graph.websiteFileLocation)
self.graphReports.append(graph)
if reportDataObject.equation.fittingTarget[-3:] != "REL":
graph = ReportsAndGraphs.AbsoluteErrorVsDependentData_ScatterPlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
self.graphReports.append(graph)
graphs.append(graph.websiteFileLocation)
else:
graph = ReportsAndGraphs.RelativeErrorVsDependentData_ScatterPlot(reportDataObject)
graph.rank = i + self.rank
graph.PrepareForReportOutput()
graphs.append(graph.websiteFileLocation)
self.graphReports.append(graph)
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.6: i=", str(i), 'equation name =', reportDataObject.equation.GetDisplayName()
dataForOneEquation = {}
splitted = listItem[1].split('.')
dataForOneEquation['moduleName'] = splitted[-1]
dataForOneEquation['displayName'] = reportDataObject.equation.GetDisplayName()
dataForOneEquation['URLQuotedModuleName'] = urllib.quote(splitted[-1])
dataForOneEquation['URLQuotedDisplayName'] = urllib.quote(reportDataObject.equation.GetDisplayName())
dataForOneEquation['displayHTML'] = reportDataObject.equation.GetDisplayHTML()
dataForOneEquation['graphWebSiteLocations'] = graphs
dataForOneEquation['rank'] = i + self.rank
dataForOneEquation['dimensionality'] = self.dimensionality
dataForOneEquation['fittingTarget'] = reportDataObject.equation.fittingTarget
dataForOneEquation['fittingTargetValue'] = targetValue
if reportDataObject.fittingTarget[-3:] != "REL": # only non-relative error fits get RMSE displayed
dataForOneEquation['rmseString'] = '<br>RMSE: ' + str(reportDataObject.equation.rmse) + '<br>'
else:
dataForOneEquation['rmseString'] = '<br>'
self.equationDataForDjangoTemplate.append(dataForOneEquation)
if self.debug: print "**** FF Results GenerateListOfOutputReports 2.7: i=", str(i), 'equation name =', reportDataObject.equation.GetDisplayName()
def __init__(self, inFittingTarget = 'SSQABS', inExtendedVersionName = 'Default'):
self.dataCache = pyeq2.dataCache()
self._dimensionality = 3
equation.Solve()
# raw data for scatterplot
x_data = equation.dataCache.allDataCacheDictionary['IndependentData'][0]
y_data = equation.dataCache.allDataCacheDictionary['IndependentData'][1]
z_data = equation.dataCache.allDataCacheDictionary['DependentData']
# create X, Y, Z mesh grid for surface plot
print "Creating mesh grid data"
xModel = numpy.linspace(min(x_data), max(x_data), 20)
yModel = numpy.linspace(min(y_data), max(y_data), 20)
X, Y = numpy.meshgrid(xModel, yModel)
tempcache = equation.dataCache
equation.dataCache = pyeq2.dataCache()
equation.dataCache.allDataCacheDictionary['IndependentData'] = numpy.array([X, Y])
equation.dataCache.FindOrCreateAllDataCache(equation)
Z = equation.CalculateModelPredictions(equation.solvedCoefficients, equation.dataCache.allDataCacheDictionary)
equation.dataCache = tempcache
# matplotlib specific code for the plots
fig = plt.figure(figsize=(float(graphWidth ) / 100.0, float(graphHeight ) / 100.0), dpi=100)
ax = fig.gca(projection='3d')
print "Creating matplotlib graph objects"
# create a surface plot using the X, Y, Z mesh data created above
ax.plot_surface(X, Y, Z, rstride=1, cstride=1, cmap=cm.coolwarm,
linewidth=1, antialiased=True)
def serialWorker(inputQueue, outputQueue):
for func, args in iter(inputQueue.get, 'STOP'): # iter() has different behaviors depending on number of parameters
outputQueue.put(func(*args))
inputQueue.task_done()
if inputQueue.unfinished_tasks == 0:
break
os.nice(10) ####################### I use this during development
global globalDataCache
globalDataCache = pyeq2.dataCache()
global globalReducedDataCache
globalReducedDataCache = {}
global globalRawData
globalRawData = '''
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
def __init__(self,
inFittingTarget='SSQABS',
inExtendedVersionName='Default'):
if inExtendedVersionName == '':
inExtendedVersionName = 'Default'
if inFittingTarget not in self.fittingTargetDictionary.keys():
raise Exception, str(
inFittingTarget
) + ' is not in the IModel class fitting target dictionary.'
self.extendedVersionHandler = eval(
'pyeq2.ExtendedVersionHandlers.ExtendedVersionHandler_' +
inExtendedVersionName.replace(' ', '') +
'.ExtendedVersionHandler_' +
inExtendedVersionName.replace(' ', '') + '()')
self.dataCache = pyeq2.dataCache()
self.upperCoefficientBounds = []
self.lowerCoefficientBounds = []
self.estimatedCoefficients = []
self.fixedCoefficients = []
self.solvedCoefficients = []
self.polyfunctional2DFlags = []
self.polyfunctional3DFlags = []
self.xPolynomialOrder = None
self.yPolynomialOrder = None
self.rationalNumeratorFlags = []
self.rationalDenominatorFlags = []
self.fittingTarget = inFittingTarget
self.independentData1CannotContainZeroFlag = False
self.independentData1CannotContainPositiveFlag = False
self.independentData1CannotContainNegativeFlag = False
self.independentData2CannotContainZeroFlag = False
self.independentData2CannotContainPositiveFlag = False
self.independentData2CannotContainNegativeFlag = False
try:
if self._dimensionality == 2:
self.exampleData = '''
X Y
5.357 0.376
5.457 0.489
5.797 0.874
5.936 1.049
6.161 1.327
6.697 2.054
6.731 2.077
6.775 2.138
8.442 4.744
9.769 7.068
9.861 7.104
'''
else:
self.exampleData = '''
X Y Z
3.017 2.175 0.320
2.822 2.624 0.629
2.632 2.839 0.950
2.287 3.030 1.574
2.207 3.057 1.725
2.048 3.098 2.035
1.963 3.115 2.204
1.784 3.144 2.570
1.712 3.153 2.721
2.972 2.106 0.313
2.719 2.542 0.643
2.495 2.721 0.956
2.070 2.878 1.597
1.969 2.899 1.758
1.768 2.929 2.088
1.677 2.939 2.240
1.479 2.957 2.583
1.387 2.963 2.744
2.843 1.984 0.315
2.485 2.320 0.639
2.163 2.444 0.954
1.687 2.525 1.459
1.408 2.547 1.775
1.279 2.554 1.927
1.016 2.564 2.243
0.742 2.568 2.581
0.607 2.571 2.753
'''
except:
pass