Table 5
Confusion matrices for the neural network shown in Fig. 20.
| Without polarization | ||||
|---|---|---|---|---|
| Actual | Ocean | Desert | Veget. | Clouds |
| Pred | ||||
| Ocean | 0.90 | 0.06 | 0.03 | 0.05 |
| Desert | 0.05 | 0.85 | 0.08 | 0.06 |
| Veget. | 0.02 | 0.05 | 0.85 | 0.05 |
| Clouds | 0.03 | 0.04 | 0.04 | 0.84 |
| With polarization | ||||
|---|---|---|---|---|
| Actual | Ocean | Desert | Veget. | Clouds |
| Pred | ||||
| Ocean | 0.92 | 0.06 | 0.03 | 0.05 |
| Desert | 0.04 | 0.85 | 0.08 | 0.05 |
| Veget. | 0.01 | 0.05 | 0.85 | 0.04 |
| Clouds | 0.03 | 0.03 | 0.04 | 0.85 |
Notes. The accuracy of the retrieval using directional light curves without polarization (top), and with polarization (bottom), for planets with the best configuration from Fig. 17, without noise. Shown are the fractions of each surface types’ facets (columns) that are classified as a specific type (rows). The diagonals show the correct classification of each surface type (cf. Fig. 21 for Nmax = ∞). An increase or decrease with respect to the matrix without polarization (top) is marked in green and red in the matrix with polarization (bottom), respectively.
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.