A piecewise asymmetric variance penalty added to EVCL yields marginal accuracy gains on MNIST-family benchmarks, with several ties and no reported uncertainty.
Variational contin- ual learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Adaptive Variance-Penalized Continual Learning with Fisher Regularization
A piecewise asymmetric variance penalty added to EVCL yields marginal accuracy gains on MNIST-family benchmarks, with several ties and no reported uncertainty.