PPAP scores parameter importance from activity during the final training plateau instead of the whole run, and scaling optimizer updates with this profile reduces catastrophic forgetting on CIFAR-10 and CIFAR-100 relative to SI and EWC.
Continual learning in neural networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Catastrophic Forgetting Mitigation Through Plateau Phase Activity Profiling
PPAP scores parameter importance from activity during the final training plateau instead of the whole run, and scaling optimizer updates with this profile reduces catastrophic forgetting on CIFAR-10 and CIFAR-100 relative to SI and EWC.