Free Access

Table 6.

Results for a neural network trained on LSR0, LSR1, and LSR2 of m12i using only 5D kinematics as inputs.

Trainable layers χ2 [×10−2]
No transfer learning 1.07

VM 1st 2.13
Last 1.20
1st and last 2.28

ZM 1st 1.04
Last 1.02
1st and last 1.03

Notes. We compare the results with and without transfer learning on m12f LSR0. When the transfer learning is performed, either the first layer is updated, or the last layer is updated, or both. In the transfer learning step, all truth labels are derived using either the VM or ZM selections, as defined in Sect. 3. ∑χ2 quantifies how similar the vR, vϕ, and vz are to the truth-level distributions, as illustrated in Fig. 8. For each transfer learning method, we determine the optimal cut on the network score, as described in the text. In general, the ZM selection with transfer learning performed by only varying the last layer does the best job at reproducing the kinematic distributions of the true accreted stars, that is, it achieves the lowest ∑χ2.

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.