Table of Links
3. SkyCURTAINs Method and 3.1 CurtainsF4F
5. Conclusion, Acknowledgments, Data Availability, and References
APPENDIX A: CurtainsF4F TRAINING AND HYPERPARAMETER TUNING DETAILS
A1. CurtainsF4F features preprocessing
4.1 Metrics
We now demonstrate the performance of the SkyCURTAINs method on GDR2 data. To quantify the discovery potential of SkyCURTAINs method, we measure the Significance Improvement Characteristic (SIC) curve for the GD-1 stream. In Figure 7, we show the SIC curve as a function of the signal efficiency for the GD-1 stream in one of the 21 patches. This metric is defined as the ratio of the signal efficiency to the square root of the background efficiency, and essentially quantifies the improvement in the discovery significance of the signal from the method. SkyCURTAINs achieves a maximum significance improvement of ∼ 10 at ∼ 50% signal efficiency. Although a direct comparison with Via Machinae is difficult on account of different SR being used for the analysis, one can look at the maximum value of the SIC as a heuristic measure, which are comparable for both methods.
We track two other metrics to quantify the performance of SkyCURTAINs: purity 𝑝: The fraction of candidate CurtainsF4F
stars that overlap with the PWB18 identified GD-1 stream members; and signal efficiency, 𝜖𝑆 which is the fraction of GD-1 stream members that have been flagged as candidates by CurtainsF4F step. Figure 5 (right) shows the candidates from the CurtainsF4F step in the 𝜙-𝜆 space that corresponds to the GD-1 stream with a 𝑝 = 75% and 𝜖𝑆 = 36.82%. We note that it also predicts a few stars that do not form a line like structure in the 𝜙-𝜆 space. This is expected, as this stage is designed to flag any overdensity in the feature space as a potential signal candidate. To filter out the line like overdensities we perform a Hough transform on the output of CurtainsF4F step. After applying the Hough filter, the purity is improved to 91.79%, albeit at the cost of a slightly reduced signal efficiency of 34.3%.
Authors:
(1) Debajyoti Sengupta, Département de physique nucléaire et corpusculaire, University of Geneva, Switzerland ([email protected]);
(2) Stephen Mulligan, Département de physique nucléaire et corpusculaire, University of Geneva, Switzerland;
(3) David Shih, NHETC, Dept. of Physics and Astronomy, Rutgers, Piscataway, NJ 08854, USA;
(4) John Andrew Raine,, Département de physique nucléaire et corpusculaire, University of Geneva, Switzerland;
(5) Tobias Golling, Département de physique nucléaire et corpusculaire, University of Geneva, Switzerland.
This paper is
