Table 1
DLNN architecture used as an SED emulator.
| Layer | Properties | |
|---|---|---|
| Input | Nin neurons(a) | |
| Linear, fully connected | 2000 neurons | ReLU activation |
| Dropout | Dropout rate 0.35 | |
| Linear, fully connected | 1000 neurons | ReLU activation |
| Linear, fully connected | 500 neurons | ReLU activation |
| Dropout | Dropout rate 0.35 | |
| Output, fully connected | Nout neurons(b) | ReLU activation |
Notes. The input and output layers have a number of neurons that depends on the particular task (i.e., predicting fluxes in the MIPS, PACS or SPIRE bands). (a)To predict the 24 μm flux, the number of neurons in the input layer Nin corresponds to the number of the multi-band photometry (satisfying the S/N requirements) in COSMOS2020 and the 16 colours as explained in Sect. 2.2.2. To predict the PACS and SPIRE fluxes, Nin also includes the deblended far-IR/sub-mm fluxes in the previous step(s). (b)To predict the 24 μm flux, the output is a single number. To predict the PACS and SPIRE fluxes, the output consists of two and three numbers, 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.