Binary normalized layers with all parameters in {0,1} reach near-32-bit accuracy on Food-101 and WikiText-103, with some caveats.
Deep Learning
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
-
1 bit is all we need: binary normalized neural networks
Binary normalized layers with all parameters in {0,1} reach near-32-bit accuracy on Food-101 and WikiText-103, with some caveats.