
Scissorhands presents a novel machine unlearning approach that effectively erases the influence of data from trained models by identifying critical parameters and reinitializing them. The algorithm balances forgetting data while preserving model utility through a gradient projection-based approach.


coremsg

scissorhands-machine-unlearning-algorithm-for-data-influence-erasure


Scissorhands: Machine Unlearning Algorithm for Data Influence Erasure


title_rewrite


Scissorhands erases data influence in models through connection sensitivity and gradient projection.
