h0rton.h0_inference.mcmc_utils

Module Contents

Classes

HybridBNNPenalty

Wrapper for subclasses of BaseGaussianNLL that allows MCMC methods to appropriately penalize parameter samples

Functions

get_lens_kwargs(init_dict, null_spread=False)	Get the SPEMD, SHEAR kwargs with the provided init values and some conservative sigma, fixed, lower, and upper
get_ps_kwargs(measured_img_ra, measured_img_dec, astrometry_sigma, hard_bound=30.0)	Get the point source kwargs for the image positions
get_ps_kwargs_src_plane(init_dict, astrometry_sigma, hard_bound=5.0)	Get the point source kwargs for the source plane
get_light_kwargs(init_R, null_spread=False)	Get the usual sigma, lower, upper, and fixed kwargs for a SERSIC_ELLIPSE light with only the R_sersic allowed to vary
get_special_kwargs(n_img, astrometry_sigma, delta_pos_hard_bound=5.0, D_dt_init=5000.0, D_dt_sigma=1000.0, D_dt_lower=0.0, D_dt_upper=15000.0)	Get the point source kwargs for the image positions
postprocess_mcmc_chain(kwargs_result, samples, kwargs_model, fixed_lens_kwargs, fixed_ps_kwargs, fixed_src_light_kwargs, fixed_special_kwargs, kwargs_constraints, kwargs_fixed_lens_light=None, verbose=False, forward_modeling=False)	Postprocess the MCMC chain for making the chains consistent with the optimized lens model and converting parameters
get_idx_for_params(out_dim, Y_cols, params_to_remove, likelihood_class, debug=False)	Get columns corresponding to certain parameters from the BNN output
remove_parameters_from_pred(orig_pred, remove_idx, return_as_tensor=True, device=’cpu’)	Remove columns corresponding to certain parameters from the BNN output
split_component_param(string, sep=’_’, pos=2)	Split the component, e.g. lens_mass, from the parameter, e.g. center_x, from the paramter names under the Baobab convention
dict_to_array(Y_cols, kwargs_lens, kwargs_source, kwargs_lens_light=None, kwargs_ps=None)	Reformat kwargs into np array. Used to feed the current iteration of MCMC kwargs into the BNN posterior evaluation.

h0rton.h0_inference.mcmc_utils.get_lens_kwargs(init_dict, null_spread=False)

Get the SPEMD, SHEAR kwargs with the provided init values and some conservative sigma, fixed, lower, and upper

The sigma, fixed, lower, and upper kwargs are hardcoded.

init_dict : dict

the init values for each of the parameters in SPEMD and SHEAR

h0rton.h0_inference.mcmc_utils.get_ps_kwargs(measured_img_ra, measured_img_dec, astrometry_sigma, hard_bound=30.0)

Get the point source kwargs for the image positions

measured_img_ra : np.array

measured ra of the images

measured_img_dec : np.array

measured dec of the images

astrometry_sigma : float

astrometric uncertainty in arcsec

hard_bound : float

hard bound of the image positions around zero in arcsec

list of dict

list of init, sigma, fixed, lower, and upper kwargs

h0rton.h0_inference.mcmc_utils.get_ps_kwargs_src_plane(init_dict, astrometry_sigma, hard_bound=5.0)

Get the point source kwargs for the source plane

measured_img_ra : np.array

measured ra of the images

measured_img_dec : np.array

measured dec of the images

astrometry_sigma : float

astrometric uncertainty in arcsec

hard_bound : float

hard bound of the image positions around zero in arcsec

list of dict

list of init, sigma, fixed, lower, and upper kwargs

h0rton.h0_inference.mcmc_utils.get_light_kwargs(init_R, null_spread=False)

Get the usual sigma, lower, upper, and fixed kwargs for a SERSIC_ELLIPSE light with only the R_sersic allowed to vary

h0rton.h0_inference.mcmc_utils.get_special_kwargs(n_img, astrometry_sigma, delta_pos_hard_bound=5.0, D_dt_init=5000.0, D_dt_sigma=1000.0, D_dt_lower=0.0, D_dt_upper=15000.0)

Get the point source kwargs for the image positions

measured_img_ra : np.array

measured ra of the images

measured_img_dec : np.array

measured dec of the images

astrometry_sigma : float

astrometric uncertainty in arcsec

hard_bound : float

hard bound of the image positions around zero in arcsec

list of dict

list of init, sigma, fixed, lower, and upper kwargs

h0rton.h0_inference.mcmc_utils.postprocess_mcmc_chain(kwargs_result, samples, kwargs_model, fixed_lens_kwargs, fixed_ps_kwargs, fixed_src_light_kwargs, fixed_special_kwargs, kwargs_constraints, kwargs_fixed_lens_light=None, verbose=False, forward_modeling=False)

Postprocess the MCMC chain for making the chains consistent with the optimized lens model and converting parameters

pandas.DataFrame

processed MCMC chain, where each row is a sample

class h0rton.h0_inference.mcmc_utils.HybridBNNPenalty(Y_cols, likelihood_class, mcmc_train_Y_mean, mcmc_train_Y_std, exclude_vel_disp, device)

Wrapper for subclasses of BaseGaussianNLL that allows MCMC methods to appropriately penalize parameter samples

set_bnn_post_params(self, bnn_post_params)

Set BNN posterior parameters, which define the penaty function

evaluate(self, kwargs_lens, kwargs_source, kwargs_lens_light=None, kwargs_ps=None, kwargs_special=None, kwargs_extinction=None)

h0rton.h0_inference.mcmc_utils.get_idx_for_params(out_dim, Y_cols, params_to_remove, likelihood_class, debug=False)

Get columns corresponding to certain parameters from the BNN output

out_dim : int Y_cols : list of str

remove_idx : list

indices of the columns removed in orig_pred

remove_param_idx : list

indices of the parameters in Y_cols

h0rton.h0_inference.mcmc_utils.remove_parameters_from_pred(orig_pred, remove_idx, return_as_tensor=True, device='cpu')

Remove columns corresponding to certain parameters from the BNN output

orig_pred : np.array of shape [n_lenses, out_dim]

the BNN output

orig_Y_cols : list of str

the original list of Y columns

params_to_remove : list of str

list of colums to remove. They must belong to orig_Y_cols

new_pred : np.array of shape [n_lenses, out_dim - len(params_to_remove)]

h0rton.h0_inference.mcmc_utils.split_component_param(string, sep='_', pos=2)

Split the component, e.g. lens_mass, from the parameter, e.g. center_x, from the paramter names under the Baobab convention

string : str

the Baobab parameter name (column name)

sep : str

separation character between the component and the parameter. Default: ‘_’ (Baobab convention)

pos : int

position of the component when split by the separation character. Default: 2 (Baobab convention)

tuple of str

component, parameter

h0rton.h0_inference.mcmc_utils.dict_to_array(Y_cols, kwargs_lens, kwargs_source, kwargs_lens_light=None, kwargs_ps=None)

Reformat kwargs into np array. Used to feed the current iteration of MCMC kwargs into the BNN posterior evaluation.