Table 3.
Summary of the performance of the CNN at identifying edge-on lens candidates.
| Grade | Mag. bin | Candidates | Candidates | Completeness | Purity of the | Prevalence in | Ratio of CNN-purity |
|---|---|---|---|---|---|---|---|
| found | expected | CNN sample | parent sample | to base prevalence | |||
| A + B | 17 to 18 | 4 | 11 | 36% | 7.7 × 10−3 | 3.3 × 10−5 | 231.2 |
| A + B | 18 to 19 | 7 | < 37 | > 19% | 1.9 × 10−3 | 0 | |
| A + B | 19 to 20 | 10 | < 108 | > 9% | 1.1 × 10−3 | 0 | |
| A + B | 20 to 20.5 | 0 | < 106 | 0 | 0 | 0 | |
| C | 17 to 18 | 9 | 55 | 16% | 1.7 × 10−2 | 1.7 × 10−4 | 104.0 |
| C | 18 to 19 | 21 | 100 | 21% | 5.6 × 10−3 | 1.0 × 10−4 | 56.2 |
| C | 19 to 20 | 11 | 95 | 12% | 1.2 × 10−3 | 3.3 × 10−5 | 35.7 |
| C | 20 to 20.5 | 1 | < 106 | > 1% | 2.5 × 10−4 | 0 | |
Notes. Metrics are marginalised by grade and magnitude bin to enable direct comparison with Table 1. We present the number of candidates found, the expected number of candidates based on Table 1, and the completeness (number found / expected number). We also include the fraction of lens candidates among sources inspected for the CNN-selected sample and for the randomly selected sample, respectively; and the ratio between these two, corresponding to the boost in lensing prevalence due to the CNN selection relative to the random selection.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.