src.superphot_plus.utils

Module Contents

Functions

get_band_extinctions_from_mwebv(mwebv, wvs)	Get extinction list from MWEBV value and wavelengths.
get_band_extinctions(ra, dec, wvs)	Get g- and r-band extinctions in magnitudes for a single
calc_accuracy(pred_classes, test_labels)	Calculates the accuracy of the random forest after predicting all
f1_score(pred_classes, true_classes[, class_average])	Calculates the F1 score for the classifier. If
convert_mags_to_flux(mag, merr, zero_point)	Converts magnitudes to flux.
flux_model(cube, t_data, b_data, max_flux, ...)	Given "cube" of fit parameters, returns the flux measurements for
params_valid(beta, gamma, tau_rise, tau_fall)	Check if parameters are valid given certain model constraints.
villar_fit_constraint(x)
get_numpyro_cube(params, max_flux, ref_band, ordered_bands)	Convert output param dict from numpyro sampler to match that
calculate_log_likelihood(cube, lightcurve, ...)	Calculate the log-likelihood of a single lightcurve.
calculate_mse(cube, lightcurve, unique_bands, ref_band)	Calculate the mean-square error for a lightcurve.
calculate_chi_squareds(cubes, t, f, ferr, b, max_flux)	Gets the negative chi-squared of posterior fits from the model
create_dataset(features, labels)	Creates a PyTorch dataset object from numpy arrays.
calculate_accuracy(y_pred, y)	Calculate the prediction accuracy.
epoch_time(start_time, end_time)	Sets the time it takes for each epoch to train.
save_test_probabilities(obj_names, pred_probabilities, ...)	Saves probabilities to a separate file for ROC curve generation.
write_metrics_to_file(config, true_classes, ...)	Calculates the accuracy and f1 score metrics for the
extract_wrong_classifications(true_classes, ...)	Extracts the wrong model classifications and copies them to a separate folder.
get_session_metrics(metrics)	Calculates the validation loss and accuracy for the hyperparameter set.
log_metrics_to_tensorboard(metrics, config, trial_id)	Calculates the training and validation accuracies and losses

get_band_extinctions_from_mwebv(mwebv, wvs)

Get extinction list from MWEBV value and wavelengths.

get_band_extinctions(ra, dec, wvs)

Get g- and r-band extinctions in magnitudes for a single supernova lightcurve based on right ascension (RA) and declination (DEC).

Parameters:

• ra (float) – The right ascension of the object of interest, in degrees.

• dec (float) – The declination of the object of interest, in degrees.

• wvs (list or np.ndarray) – Array of wavelengths, in angstroms.

Returns:

ext_dict – A dictionary mapping bands to extinction magnitudes for the given coordinates.

Return type:

Dict

calc_accuracy(pred_classes, test_labels)

Calculates the accuracy of the random forest after predicting all classes.

Parameters:

• pred_classes (np.ndarray of int) – Classes predicted by MLP.

• test_labels (np.ndarray of int) – True spectroscopic classes.

Returns:

Accuracy – The percentage of the predictions that are correct.

Return type:

float

Raises:

ValueError – If the pred_classes or test_labels arrays are empty or if they are of mismatched sizes.

f1_score(pred_classes, true_classes, class_average=False)

Calculates the F1 score for the classifier. If class_average=True, then the macro-F1 is used. Else, uses the weighted-F1 score.

Parameters:

• pred_classes (np.ndarray of int) – Classes predicted by MLP.

• true_classes (np.ndarray of int) – True spectroscopic classes.

• class_average (bool, optional) – Determines whether F1 score is weighted equally for each class, or by number of samples per class. Defaults to False.

Returns:

The calculated F1 score.

Return type:

float

convert_mags_to_flux(mag, merr, zero_point)

Converts magnitudes to flux.

Parameters:

• mag (array-like) – The magnitudes.

• merr (array-like) – The error in magnitudes.

• zero_point (float) – The zero point in the magnitude system.

Returns:

fluxes, flux_unc – The calculated fluxes and their uncertainties.

Return type:

tuple of numpy array

flux_model(cube, t_data, b_data, max_flux, ordered_bands, ref_band)

Given “cube” of fit parameters, returns the flux measurements for a given set of time and band data.

Parameters:

• cube (array-like) – The cube of fit parameters.

• t_data (array-like) – The time data.

• b_data (array-like) – The band data.

• ordered_bands (array-like) – The band names in the cube’s parameter order.

• ref_band (str) – The base/reference band which all other values are multiples of.

Returns:

f_model – The flux model for the given set of time and band data.

Return type:

numpy array

params_valid(beta, gamma, tau_rise, tau_fall)

Check if parameters are valid given certain model constraints.

Parameters:

• beta (float) – Parameter beta.

• gamma (float) – Parameter gamma.

• tau_rise (float) – Parameter tau_rise.

• tau_fall (float) – Parameter tau_fall.

Returns:

True if parameters are valid, False otherwise.

Return type:

bool

villar_fit_constraint(x)

get_numpyro_cube(params, max_flux, ref_band, ordered_bands)

Convert output param dict from numpyro sampler to match that of dynesty.

Parameters:

• params (dict) – Parameter dictionary

• max_flux (float) – Max flux of light curve

• aux_bands (array-like, optional) – The names of auxiliary bands, in order. If None or excluded, attempts to infer them from the dictionary.

Returns:

• cube (np.ndarray) – Array of all equal-weight parameter draws

• aux_bands (np.ndarray) – Auxiliary bands, including those inferred if input arg was None.

calculate_log_likelihood(cube, lightcurve, unique_bands, ref_band)

Calculate the log-likelihood of a single lightcurve. Copied from dynesty_sampler.py

Parameters:

• cube (np.ndarray) – Array of parameters. Must be length 7 * B where B is the number of unique bands.

• lightcurve (Lightcurve) – The lightcurve object to evaluate.

• unique_bands (list) – A list of bands to use.

• ref_band (str) – The reference band.

Returns:

logL – Log-likelihood value.

Return type:

float

calculate_mse(cube, lightcurve, unique_bands, ref_band)

Calculate the mean-square error for a lightcurve.

Parameters:

• cube (np.ndarray) – Array of parameters. Must be length 7 * B where B is the number of unique bands.

• lightcurve (Lightcurve) – The lightcurve object to evaluate.

• unique_bands (list) – A list of bands to use.

• ref_band (str) – The reference band.

Returns:

mse – The mean square error of the predictions

Return type:

float

calculate_chi_squareds(cubes, t, f, ferr, b, max_flux, ordered_bands=None, ref_band='r')

Gets the negative chi-squared of posterior fits from the model parameters and original data files. :param names: The names of the objects. :type names: list of str :param fit_dir: The directory where the fit files are located. :type fit_dir: str :param data_dirs: The directories where the data files are located. :type data_dirs: list of str :param ordered_bands: Bands in order they appear in cubes. Defaults to ZTF band order. :type ordered_bands: list of str :param ref_band: Base/reference band. Defaults to ‘r’. :type ref_band: str

Returns:

log_likelihoods – The log likelihoods for each object.

Return type:

np.ndarray

create_dataset(features, labels)

Creates a PyTorch dataset object from numpy arrays.

Parameters:

• features (np.ndarray) – The features array.

• labels (np.ndarray) – The labels array.

• idxs (np.ndarray, optional) – The indices array. Defaults to None.

Returns:

The created dataset.

Return type:

torch.utils.data.TensorDataset

calculate_accuracy(y_pred, y)

Calculate the prediction accuracy.

Parameters:

• y_pred (torch.Tensor) – The predicted tensor.

• y (torch.Tensor) – The true tensor.

Returns:

The calculated accuracy.

Return type:

torch.Tensor

epoch_time(start_time, end_time)

Sets the time it takes for each epoch to train.

Parameters:

• start_time (float) – The start time.

• end_time (float) – The end time.

Returns:

A tuple containing the elapsed minutes and elapsed seconds.

Return type:

tuple

save_test_probabilities(obj_names, pred_probabilities, output_path, true_labels=None, target_label=None)

Saves probabilities to a separate file for ROC curve generation.

Parameters:

• output_filename (str) – The file name to save to.

• pred_probabilities (array-like) – The prediction probabilities.

• true_label (str or int) – The true label.

• output_dir (str) – Where to store the generated file.

write_metrics_to_file(config, true_classes, pred_classes, prob_above_07)

Calculates the accuracy and f1 score metrics for the test set and outputs them to a log file.

Parameters:

• config (ModelConfig) – The configuration of the model used for evaluation.

• true_classes (np.ndarray) – The ground truth for the test ZTF objects.

• pred_classes (np.ndarray) – The predicted classes for the test ZTF objects.

• prob_above_07 (np.ndarray) – Indicates which predictions had a 70% confidence.

• log_file (str) – The file where the metrics information will be written.

extract_wrong_classifications(true_classes, pred_classes, ztf_test_names, fit_folder, wc_folder)

Extracts the wrong model classifications and copies them to a separate folder.

Parameters:

• true_classes (np.ndarray) – The ground truth for the classified samples.

• predicted_classes (np.ndarray) – The predictions for the classified samples.

• ztf_test_names (np.ndarray) – The ZTF object names for the classified samples.

get_session_metrics(metrics)

Calculates the validation loss and accuracy for the hyperparameter set. The best model is considered the one with the lowest validation loss.

Parameters:

metrics (tuple) – Tuple containing the training accuracies and losses, and the validation accuracies and losses, for each epoch and fold.

Returns:

The mean validation loss and accuracy for the hyperparameter set.

Return type:

tuple

log_metrics_to_tensorboard(metrics, config, trial_id, base_dir='runs')

Calculates the training and validation accuracies and losses for each epoch (by averaging each fold) and logs these metrics to Tensorboard using a SummaryWriter. It also stores the run configuration for further reference.

Parameters:

• metrics (tuple) – Tuple containing the training accuracies and losses, and the validation accuracies and losses, for each epoch and fold.

• config (ModelConfig) – The model’s training configuration.

• trial_id (str) – The experiment identifier.

• base_dir (str) – The directory where all tensorboard metrics should be stored. Defaults to “runs”.

Returns:

The training losses and accuracies, followed by the validation losses and accuracies, for each epoch.

Return type:

tuple