Fig. 5: Comparison between the two schemes for neural reuse across multiple downstream tasks.

a Processes of two schemes for neural reuse in multiple downstream tasks. Arrows with solid line represent neural reuse on the task flow, and arrows with dotted line represent the ion relaxation process. b Boxplots generated by the loss values calculated on the corresponding dataset. It reflects the measurement level of experimental data, which repeatedly recorded loss in all subsamples received randomly. c Learning curves within 20 epochs for each task. It shows the evolution trend of the learning speed and accuracy in each task flow under the two schemes. d RMSE for two schemes. The RMSE on each scheme downstream task is calculated on its matched test set to reveal the effect after transferring features in varied schemes.