Skip to main content
SearchLogin or Signup

Random Forests for Systematics Removal in Spitzer IRAC Light Curves

Published onJun 01, 2020
Random Forests for Systematics Removal in Spitzer IRAC Light Curves

We present a new method employing machine learning techniques for correcting systematics in IRAC high precision photometry based on Random Forest techniques. The goal of this work is to use machine learning to measure the shape of the intrinsic intrapixel response function. Because the intrapixel response function is a regression problem and not a simple function of any single parameter, we use the power of supervised machine learning to do a multidimensional regression using a rich calibration dataset. We aim to use the large amount of publicly available calibration data for the sweet spot pixel to make a fast, easy to use, accurate correction to science data. This correction on calibration data has the advantage of using an independent dataset to correct science data instead of using the science data on itself which has the disadvantage of including astrophysical variations. We show results of the abilities of boosted random forests to reduce data and measure eclipse depths with archival exoplanet data which has also been analyzed by current best methods in the literature. We have produced a model which can measure the ten eclipse observations of XO3-b to be 1483+- 200 parts per million. This is a comparable average depth to those in the literature (as measured during a 2016 IRAC data challenge), however the spread in measurements is something like 30-100% larger than those literature values, depending on the reduction met


No comments here