UCLA Machine Learning Reading Group: Planets ML Project

This repository is a fork of https://github.com/tensorflow/models, using the astronet part of the Tensorflow model library for training on light curves from the Kepler space telescope.

astronet README - contains instructions on replicating Shallue et al. (2018)

Setting up environment on SciServer

If using the current (2018-04-23) python DEV docker container.

Packages installed with pip:

• pip install tensorflow==1.5
• pip install numpy --upgrade
• pip install pydl

Packages installed with conda:

• conda install pandas bazel astropy

Notes on replicating the astronet code

When downloading the TCE file, be sure to specifically select the following. By default, the av_training_set column is not selected. You can select it using the select columns button on the top left of the webpage.

• rowid: Integer ID of the row in the TCE table.
• kepid: Kepler ID of the target star.
• tce_plnt_num: TCE number within the target star.
• tce_period: Period of the detected event, in days.
• tce_time0bk: The time corresponding to the center of the first detected event in Barycentric Julian Day (BJD) minus a constant offset of 2,454,833.0 days.
• tce_duration: Duration of the detected event, in hours.
• av_training_set: Autovetter training set label; one of PC (planet candidate), AFP (astrophysical false positive), NTP (non-transiting phenomenon), UNK (unknown).

https://exoplanetarchive.ipac.caltech.edu/cgi-bin/TblView/nph-tblView?app=ExoTbls&config=q1_q17_dr24_tce

NOTE: the column rowid is actually called loc_rowid in the .csv file. So either change that to rowid or change the code to use loc_rowid in generate_input_records.py