Reducing unlearning costs via low influence point identification

A new method reduces the computational cost of machine learning unlearning by identifying and targeting only high-influence data points. By leveraging influence functions, the approach avoids the overhead of treating all forget-set points as equally necessary to remove.