Table 1 PCC and R2 score of the output features for DNN-TP trained by three different methodologies.
Output parameter | Direct learning | Generic Transfer Learning | Physics-based transfer learning | |||
|---|---|---|---|---|---|---|
PCC | R2 | PCC | R2 | PCC | R2 | |
Modal Gain | 0.913 | 0.831 | 0.916 (+ 0.39%) | 0.837 (+ 0.73%) | 0.922 (+ 0.99%) | 0.850 (+ 2.19%) |
Linewidth | 0.890 | 0.789 | 0.886 (-0.48%) | 0.780 (-0.03%) | 0.891 (+ 0.11%) | 0.792 (+ 0.47%) |
Emission wavelength | 0.966 | 0.932 | 0.967 (+ 0.15%) | 0.935 (+ 0.28%) | 0.967 (+ 0.17%) | 0.935 (+ 0.36%) |
Effective injection energy | 0.984 | 0.968 | 0.984 (-0.03%) | 0.967 (-0.06%) | 0.985 (+ 0.08%) | 0.969 (+ 0.15%) |
Effective extraction energy | 0.990 | 0.979 | 0.992 (+ 0.19%) | 0.984 (+ 0.54%) | 0.992 (+ 0.17%) | 0.984 (+ 0.47%) |
Average | 0.948 | 0.900 | 0.949 (+ 0.04%) | 0.901 (+ 0.09%) | 0.951 (+ 0.29%) | 0.906 (+ 0.69%) |
