Table 3.
Performance of model training with different weight layers, self-attention, or r-radius
Weight layer | Self-attention | r-radius | MSE | |
---|---|---|---|---|
Single | No | 1 | 1.0848 | 0.8477 |
Single | No | 2 | 0.9804 | 0.8623 |
Single | No | 3 | 1.0694 | 0.8498 |
Multi | No | 1 | 1.0654 | 0.8504 |
Multi | No | 2 | 1.0078 | 0.8585 |
Multi | No | 3 | 0.9767 | 0.8628 |
Multi | Yes | 1 | 1.0724 | 0.8495 |
Multi | Yes | 2 | 0.9785 | 0.8626 |
Multi | Yes | 3 | 0.9384 | 0.8683 |
Multi | Yes | 4 | 0.9917 | 0.8608 |
Multi | Yes | 5 | 1.2083 | 0.8304 |
Bold values represents the result is the best performance among the models participating in the comparison