Core Concepts
Neuer Ansatz zur Bewältigung von katastrophalem Vergessen und Ungleichgewicht zwischen Stabilität und Plastizität in exemplarfreiem CIL.
Stats
"Extensive experiments on benchmark CIL datasets and settings demonstrate that our method excels previous exemplar-free CIL methods with non-marginal improvements and achieves state-of-the-art performance."
"Compared with SOTA exemplar-free CIL method SEED(ICLR 2024), when N = 20, our method is 9.68 percent more accurate for the average incremental accuracy Accavg and 14.24 percent more accurate for the final accuracy AccL."
"Our method greatly outperforms all data-free CIL baselines in all incremental settings."
Quotes
"Our method achieves the highest average and final accuracy among all approaches with non-marginal improvements."
"Our method greatly outperforms all baseline methods in all incremental settings."
"The results demonstrate that our method performs much better to mitigate the catastrophic forgetting problem in CIL."