To address the discrepancy modeling problem, prior works propose the local learning method to provide a customized local model trained by a small collection of local data for each user.
The gate model is trained by data with more mainstream users and thus focuses more on how to assign gate values to improve utility for mainstream users.
A more precise and unbiased gate mechanism is needed to ensure that expert models that are effective for niche users are not overlooked in assignments.
A key principle of the gate model is that the assignment of expert models should reflect their effectiveness for the target user, guided by the loss function.
[
Collection
]
[
|
...
]