Table 6.
Neural network performance for input spaces in which either skyErr or fibre E has been normalized with some other parameters that were not included as inputs (see Sect. 3.3.2).
| Input space | Accuracy | Previous accuracy |
|---|---|---|
| SkyErr as if with dark variance | 92.10 ± 0.28% | 92.04 ± 0.26% |
| SkyErr as if with counts & dark variance | 92.02 ± 0.21% | 91.98 ± 0.27% |
| SkyErr as if in fibre E – all-input-set | 91.02 ± 0.32% | 91.48 ± 0.31% |
| SkyErr pre-normalized | 90.88 ± 0.25% | 83.30 ± 0.38% |
| Fibre E as if in BCE | 79.72 ± 0.32% | 88.68 ± 0.31% |
| Fibre fractional errors | 88.60 ± 0.55% | 89.06 ± 0.42% |
| Fibre E as if in BCEn | 88.40 ± 0.33% | 88.76 ± 0.25% |
| Fibre E as if in CE | 77.66 ± 0.66% | 88.66 ± 0.37% |
| Fibre E pre-normalized | 86.74 ± 0.40% | 83.76 ± 0.32% |
| SkyErr pre-normalized logarithmic scale | 92.64 ± 0.15% | ... |
Notes. The last column gives the accuracy for the input space prior to the applied modification. For the skyErr and fibre E pre-normalized spaces (rows 4 and 9), the last column compares their respective 5D isolated min-max value.
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.