Analytic Learning for Exemplar-Free Generalized Class Incremental Learning

Proposing G-ACIL as an exemplar-free solution for Generalized Class Incremental Learning, achieving weight-invariant property and superior performance.