def __init__(self, g_parameters, image_renderer, snr):
self.g_parameters = g_parameters
self.snr = snr
self.model = galfun.getGalaxiesModels(g_parameters=self.g_parameters)
self.image_renderer = image_renderer
#we do not want to mask or crop the images used to obtain the partials.
self.image_renderer_partials = galfun.ImageRenderer(stamp=self.image_renderer.stamp)
self.image = self.image_renderer.getImage(self.model)
_, self.var_noise = galfun.addNoise(self.image, self.snr, 0)
self.steps = defaults.getSteps(self.g_parameters, self.image_renderer)
self.param_names = g_parameters.ordered_fit_names
self.num_params = len(self.param_names)
self.num_galaxies = self.g_parameters.num_galaxies
self.derivatives_images = self.derivativesImages()
self.second_derivatives_images = self.secondDerivativesImages()
self.fisher_matrix_images = self.fisherMatrixImages()
self.fisher_matrix = self.fisherMatrix()
self.covariance_matrix = self.covarianceMatrix()
self.correlation_matrix = self.correlationMatrix()
self.bias_matrix_images = self.biasMatrixImages()
self.bias_matrix = self.biasMatrix()
self.bias_images = self.biasImages()
self.biases = self.getBiases()
self.fisher_condition_number = self.fisherConditionNumber()
def derivativesImages(self):
"""Return images of the partial derivatives of the galaxy.
The partial differentiation includes each of the different parameters
that describe the galaxy.
"""
partials_images = {}
for i in range(self.num_params):
param = self.param_names[i]
params_up = copy.deepcopy(self.g_parameters.params)
params_up[param] += self.steps[param]
params_down = copy.deepcopy(self.g_parameters.params)
params_down[param] -= self.steps[param]
gal_up = galfun.getGalaxiesModels(params_up)
gal_down = galfun.getGalaxiesModels(params_down)
img_up = self.image_renderer_partials.getImage(gal_up)
img_down = self.image_renderer_partials.getImage(gal_down)
partials_images[param] = ((img_up - img_down)/(2 * self.steps[param])).array
return partials_images
def secondDerivativesImages(self):
"""Return the images for the second derivatives of the given galaxy."""
secondDs_gal = {}
for i in range(self.num_params):
for j in range(self.num_params):
param_i = self.param_names[i]
param_j = self.param_names[j]
params_iup_jup = copy.deepcopy(self.g_parameters.params)
params_iup_jup[param_i] += self.steps[param_i]
params_iup_jup[param_j] += self.steps[param_j]
params_idown_jup = copy.deepcopy(self.g_parameters.params)
params_idown_jup[param_i] -= self.steps[param_i]
params_idown_jup[param_j] += self.steps[param_j]
params_iup_jdown = copy.deepcopy(self.g_parameters.params)
params_iup_jdown[param_i] += self.steps[param_i]
params_iup_jdown[param_j] -= self.steps[param_j]
params_idown_jdown = copy.deepcopy(self.g_parameters.params)
params_idown_jdown[param_i] -= self.steps[param_i]
params_idown_jdown[param_j] -= self.steps[param_j]
gal_iup_jup = galfun.getGalaxiesModels(params_iup_jup)
gal_idown_jup = galfun.getGalaxiesModels(params_idown_jup)
gal_iup_jdown = galfun.getGalaxiesModels(params_iup_jdown)
gal_idown_jdown = galfun.getGalaxiesModels(params_idown_jdown)
img_iup_jup = self.image_renderer_partials.getImage(gal_iup_jup)
img_idown_jup = self.image_renderer_partials.getImage(gal_idown_jup)
img_iup_jdown = self.image_renderer_partials.getImage(gal_iup_jdown)
img_idown_jdown = self.image_renderer_partials.getImage(gal_idown_jdown)
secondDs_gal[param_i, param_j] = ((img_iup_jup + img_idown_jdown -
img_idown_jup - img_iup_jdown)/
(4*self.steps[param_i]*self.steps[param_j])).array
return secondDs_gal
def objFunc(fit_params, image_renderer, data, variance_noise, **kwargs):
gal_model = galfun.getGalaxiesModels(fit_params=fit_params.valuesdict(),**kwargs)
model = image_renderer.getImage(gal_model)
return ((model - data).array.ravel()) / math.sqrt(variance_noise)
