FSG trains a fast MLP and a slow Mamba-based hypernetwork on current and historical gradients to approximate sign-function derivatives, reporting higher CIFAR accuracy than several baselines.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fast and Slow Gradient Approximation for Binary Neural Network Optimization
FSG trains a fast MLP and a slow Mamba-based hypernetwork on current and historical gradients to approximate sign-function derivatives, reporting higher CIFAR accuracy than several baselines.