Assessing RAIL’s Estimation of Photometric Redshifts

Alice Crafford

The primary objective of this project will be to characterize the limitations in the ability of machine learning algorithms in the Python library RAIL to estimate the redshifts of distant galaxies, using photometric data. These estimators are trained on ideal data sets, but there will always be inconsistencies between the real data that the estimator assesses and its training set. Quantifying the effects of these inconsistencies is essential to understanding the effectiveness of estimators. The Python library RAIL will be used to generate mock data containing inconsistencies mimicking those found in real data. These mock data will then be passed through RAIL’s estimators, and their performance metrics will be assessed in order to more quantitatively understand the effects that specific deviations from the algorithm’s training set have on performance. Precisely knowing the limitations of redshift measurement tools will inform what kinds of science they can be used to perform, and assist in continuing to improve their accuracy and precision. This may have significant applications to research done with data collected by the LSST, due to the very high volume of imaging data it will be gathering in the frequency bands used to estimate photometric redshifts. Using this data to study the redshifts of extragalactic objects will contribute to a better understanding of the expansion rate of the universe, and thus to the study of dark energy, cited as one of the primary science objectives of the LSST.

 

This poster will be displayed on Monday and Tuesday.

 

 

Career Stage: 
Undergrad Student