Bug fix for ITERATE option in image BACKGROUNDS

The previous version was missing one tiny change to make the ITERATE option fully functional. This release includes it.

PARAMS option in USERDISTs

Version 0.0.2.2 of deeplenstronomy has 2 main changes:

• A new PARAMS option for USERDIST column selection
• Minor bug fixes for BACKGROUNDS and time delay shift calculations

Iterative Image Backgrounds and Geometry Protections

Version 0.0.2.1 of deeplenstronomy contains 3 main updates:

1. Image backgrounds are no longer Poisson resampled.
2. Image backgrounds can be used iteratively with the ITERATE key in the configuration file
3. Simulated systems with unphysical geometries (e.g. if PLANE_2 has a larger redshift than PLANE_1) are deleted from the images and metadata

Verified stability of new time-series improvements

The new time series features introduce in versions 0.0.1.* have now been tested in several different analyses. There are no major changes in this release.

Accuracy Enhancements

This release solves two existing issues introduced by the explicit calculation of time delays in version 1.7:

• the explicit calculation of time delays required each galaxy to have a mass profile and each configuration to have at least two planes. The automated checking functionalities now make sure the configuration file meets these requirements

• nites in the cadence but far outside the SED caused inaccuracies in extrapolation, the magnitudes for these nites are now set to 99

Improved Accuracy for Time Series

This release includes several accuracy improvements in the modeling of time-dependent behavior in the simulations. Specifically:

• Improve accuracy of calculated magnitudes from SEDs (SNe, KN) by calibrating to real SN-Ia brightnesses
• Track cosmographic time delays in the metadata
• Make image backgrounds compatible with time series

As well, to conserve memory usage in large simulations, the simulation input dictionaries generated by input_reader.Organizer are now written to disk and each configuration reads only it's simulation input dictionary into memory when it is simulated

Introducing the DES Deep Fields

This release includes three new distributions: des_deep_seeing, des_deep_exposure_time, and des_deep_magnitude_zero_point for simulating des deep field quality images.

This release also contains an update to timeseries.py that fixes a bug in the K-Correction calculation.

These updates are in beta release for this version and may have additional updates / patches in the near future. Please open an issue if they cause any problems.

Bug fixes for lsst survey mode and supernovae timeseries

This release contains multiple bug fixes.

• The survey parameter in deeplenstronomy.make_dataset() caused an error due to a bug in surveys.lsst() which has now been fixed.
• Previously high redshift supernovae would cause a NaN value to enter the simulated image pixels due to an improperly calculated kcorrection. timeseries.py now has safeguards for this behavior.
• USERDISTS caused issues when simulating time series data with several nights because the distribution was not sampled enough times to have a value for every image that would be simulated. The USERDISTs are now sampled a correct number of times to prevent this behavior

Speed and Visulization Improvements

This release contains a significant speed up for timeseries functionalities that involve integrating an SED (the ia, cc, kn, and user models) by replacing an integration with a rectangular sum approximation. Time series data generation wall clock time still scales with the number of pointings, but the process happens ~1,500 times faster now.

This release also contains a new function for visualizing the metadata of a simulated dataset as a corner plot of 1D and 2D histograms.

Expanded Timeseries Functionalities

1. You can now use the static model for time series which puts a source of zero brightness in the image but enables the simulation of time series images. Useful to replicate single-epoch images of unchanging sources.
2. You can now pass the PEAK parameter to the TIMESERIES part of the a CONFIGURATION entry in the GEOMETRY section. PEAK can be set like any other parameter (fixed or drawn from a distribution) and sets the NITE at which a time series model, such as a supernova, reaches it's maximum brightness.
3. For lightcurves generated from SEDs (ia, cc, or kn models), you now get a noiseless realization of the magnitude and a measured realization of the magnitude at each time point.