A federated class-incremental learning method that uses condensed synthetic exemplars, produced by gradient and feature matching plus a shared VAE, to reduce catastrophic forgetting.
Blockchained federated learning for inter- net of things: A comprehensive survey
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Exemplar-condensed Federated Class-incremental Learning
A federated class-incremental learning method that uses condensed synthetic exemplars, produced by gradient and feature matching plus a shared VAE, to reduce catastrophic forgetting.