The squared gradient accumulator from Adam can serve as a free, drop-in approximation of the empirical Fisher diagonal across model merging, pruning, task embedding, and continual learning, with comparable performance.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fishers for Free? Approximating the Fisher Information Matrix by Recycling the Squared Gradient Accumulator
The squared gradient accumulator from Adam can serve as a free, drop-in approximation of the empirical Fisher diagonal across model merging, pruning, task embedding, and continual learning, with comparable performance.