Low-Rank Expert
A lightweight adaptation module constructed through low-rank decomposition, containing only a small number of trainable parameters. It is inserted into specific layers of a frozen backbone network to learn domain- or task-specific residual adjustments.
Timeline (last 90 days)
APRIL is an exemplar-free domain-incremental learning framework based on a shared frozen BEATs backbone that sequentially inherits and isolates low-rank experts, so that adaptation to a new domain does not alter the old models.
Low-rank experts are lightweight adapter modules constructed via low-rank decomposition (such as LoRA); each expert contains only a small number of trainable parameters and is inserted into specific layers of the frozen backbone to learn domain-specific residual adjustments.
All Facts (2)
APRIL is an exemplar-free domain-incremental learning framework based on a shared frozen BEATs backbone that sequentially inherits and isolates low-rank experts, so that adaptation to a new domain does not alter the old models.
50%UnverifiedLow-rank experts are lightweight adapter modules constructed via low-rank decomposition (such as LoRA); each expert contains only a small number of trainable parameters and is inserted into specific layers of the frozen backbone to learn domain-specific residual adjustments.
50%