emilleishida / plasticc_validation Goto Github PK

After today's Cadence White Paper telecon, we figured we'd need to make assingments of each simulated cadence to one person. @rbiswas4, @mi-dai and I will write code to do feature computation on a subset of the light curves in each cadence (stay tuned), while @daniel-muthukrishna will run his classifier on all of them (in progress).

Here are assignments (each cadence dir is ~ 40G, so it helps to have a midway account, else we have to stage it some other way).

alt_sched: @soraisam
alt_sched_rolling: @kponder
kraken_2026: @rbiswas4
kraken_2044: @gnarayan
nexus_2097: @lgalbany
rolling_10yrs_opsim: @tinapeters
rolling_mix_10yrs_opsim: @mi-dai

Now that challenge is done, should we make this repo public again? That way we can link to it in the validation and results paper.

Hi @gnarayan ,

While adding to the description of data yesterday (@emilleishida and @RickKessler , please correct/append to this), there was broad agreement on requests for the following changes. @RickKessler will probably add one to this list for correcting the galactic extinction:
It would be nice to make these changes for the next data set we send out. Please let us know what you think.

• hostgal_specz should be changed to specz. This is being supplied in the training set only, and correspond to live spectroscopy of the variable/transient, rather than host.
• mwebv_err : drop this column
• sntype should be changed to class
• photflag : drop
• passband : filter (@gnarayan shadowing objection is good! Let us hear what people think).

@RickKessler has an additional proposal:

• drop MWEBV and add a band dependent correction which is not be accurately corrected for, but mostly. 'correct the FLUXLCAL values for MWEBV based on central wavelength, and them remove MWEBV from the Kaggle challenge'

Everyone on the repo: please write about this, particularly if you have strong objections to any of these proposals, or feel free to shout on the validation channel!.