Table A.6
Efficiency of our approach.
| annotation count | Precision | Accuracy | Recall | F1 score | MCC | AP | Sum |
|---|---|---|---|---|---|---|---|
| ZTF-NEWm | |||||||
| 1000 | 94.8±1.5 | 97.4±0.6 | 96.6±1.1 | 95.7±0.9 | 93.8±1.3 | 99.4±0.3 | 577.7±2.5 |
| 1500 | 95.5±1.8 | 97.7±0.6 | 97.1±0.4 | 96.3±1.0 | 94.7±1.5 | 99.6±0.2 | 580.9±2.6 |
| 2000 | 96.8±0.9 | 98.2±0.5 | 97.1±0.8 | 97.0±0.8 | 95.6±1.1 | 99.6±0.2 | 584.3±1.9 |
| 2500 | 96.6±1.3 | 98.4±0.5 | 98.1±0.4 | 97.3±0.8 | 96.2±1.1 | 99.6±0.4 | 586.3±2.0 |
| 3000 | 98.0±0.7 | 98.8±0.2 | 97.9±0.6 | 98.0±0.3 | 97.1±0.5 | 99.9±0.1 | 589.7±1.1 |
Notes. Our methodology yields results surpassing those of the fully supervised model employing 2500 annotations, despite using only 1000 labels available. We employ precision, accuracy, recall, F1 score, Matthews correlation coefficient (MCC), and average precision (AP) as performance measurement indicators. The overall performance measurement is obtained by summing the values of these six indicators (shown in the last column).
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.