It derives a covariant, activation-free layer map for neural networks on noncompact symmetric spaces and works out explicit hyperbolic and class-switching examples.
Expressivity of deep neural networks,
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
-
Navigation through Non-Compact Symmetric Spaces: a mathematical perspective on Cartan Neural Networks
It derives a covariant, activation-free layer map for neural networks on noncompact symmetric spaces and works out explicit hyperbolic and class-switching examples.