NAP applies atom-prevalence selection to neurons, pruning 56–92% of nodes in tested feedforward networks with small accuracy loss, but its calibration claim fails on MNIST.
Title resolution pending
1 Pith paper cite this work, alongside 12 external citations. Polarity classification is still indexing.
1
Pith paper citing it
12
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Neural Feature Governance: Extending Atom Prevalence
NAP applies atom-prevalence selection to neurons, pruning 56–92% of nodes in tested feedforward networks with small accuracy loss, but its calibration claim fails on MNIST.