# PLAsTiCC: the Photometric LSST Astronomical Time-Series Classification Challenge
The PLAsTiCC dataset is a set of simulated light curves from the the Vera Rubin Observatory. It was released by the LSST Dark Energy Science Collaboration in 2018 as a data challenge on [kaggle](http://kaggle.com/c/PLAsTiCC-2018) to let data scientists from outside astronomy try their hand at classifying several kinds of objects that LSST will observe, based only on their multi-band light curves. Along with [simulated light curves](https://zenodo.org/record/2539456#.Ypi4li8RqX0), the release includes a [handy starter kit](https://github.com/LSSTDESC/plasticc-kit/) with tutorial notebooks.
(A successor data challenge, ELAsTiCC, will run later this summer.)
The light curve files contain the following columns:
```
name : description
# ------------------------------------------------------
object_id. : unique object identifier (integer)
mjd : modified julien date (float)
passband : passband integer with 0,1,2,3,4,5 --> u,g,r,i,z,y
flux : measured flux (float), corrected for Galactic extinction.
Flux zeropoint=27.5.
flux_err : uncertainty on the flux listed above (float).
detected_bool : 1 = detection from image-subtraction pipeline
```
Metadata for each object_id (ra/dec, redshifts, galactic extinction, true class in the unblinded set) can be found in the corresponding metadata file.
# ZTF nuclear sample: 10k forced photometry lightcurves from the Zwicky Transient Facility, an all-sky survey located in Palomar, California. This survey takes g- and r-band optical images all 2-3 nights of the same sky regions (as well as i-band images with a somewhat lower cadence). This makes it great for transient astronomy.
This dataset is version 1 of an effort to create a full dataset of ZTF lightcurves for transients that happen near the centers of galaxies. It will contain a lot of AGN which were flaring during ZTF observations, but also more interesting (sorry, AGN folks) transients like supernovae or tidal disruption events.
All lightcurves are available as .csv files, and are located here:
These have already been cleaned, i.e. bad datapoints have been removed.
Each csv-file contains a lot of columns, but the most interesting ones are listed here:
ampl The forced photometry flux (float)
ampl.err The error on the forced photometry flux (float)
filter The filter through which the image was taken (str, ZTF_g, ZTF_r or ZTF_i)
obsmjd The Modified Julian Day of the observation (float)
magzp The abmag zeropoint of the observation (float). With this, the forced photometry flux can be converted to a magnitude.
Additionally, there is a metadata file here:
This contains more or less secure metadata on each of the transients (e.g. classification, redshift, various computed metrics like rise time, decay time). One of the main tasks will be determining how to create training and validation samples using these metadata.
You will quickly realize that this "real world" dataset is vastly more noisy and difficult to handle in comparison to the simulated LSST data. But well, that is how the cookie crumbles ;)
An example notebook showing how to select on of these transients and plot its lightcurve is available here:
