Open ynning opened 2 months ago
May I ask how the results of the dual prediction head are combined? Is the optimal result obtained by multiplying the predicted results for each head with different weights?
As the code shows, it simply adds up the output results of the dual heads, allowing the network itself to learn and assign weights.
May I ask how the results of the dual prediction head are combined? Is the optimal result obtained by multiplying the predicted results for each head with different weights?