Open Access

Table B.1

Summary of the best hyperparameters found for the SwinV2 model across single- and multi-band cases and varying dataset sizes.

MACHO (single-band) Marker size Line width Flux errors Input format Learning rate
20 spc 5.0 1.5 True Overlay 5 ⋅ 10−5
500 spc 1.0 1.0 True Overlay 5 · 10¯6
Full dataset 1.0 2.0 True Overlay 5 · 10¯6

MACHO (two-band) Marker size Line width Flux errors Input format Learning rate

20 spc 1.0 2.0 True Grid 5 ⋅ 10−5
20 spc 2.0 0.5 True Overlay 5 ⋅ 10−5
500 spc 2.0 1.0 False Grid 5 ⋅ 10−5
500 spc 3.0 1.0 True Overlay 5 ⋅ 10−5
Full dataset 3.0 1.5 True Grid 5 ⋅ 10−6
Full dataset 1.0 2.0 True Overlay 5 ⋅ 10−6

ELAsTiCC (six-band) Marker size Line width Flux errors Input format Learning rate

Full dataset 5.0 2.0 True Grid 5 ⋅ 10−6
Full dataset 2.0 0.5 True Overlay 5 ⋅ 10−6

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.