A set of sufficient conditions and training losses for enforcing observability in neural state-space models, with one clean Mamba condition and several unproven high-probability Fourier results.
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
REJECT 1representative citing papers
citing papers explorer
-
Observability conditions for neural state-space models with eigenvalues and their roots of unity
A set of sufficient conditions and training losses for enforcing observability in neural state-space models, with one clean Mamba condition and several unproven high-probability Fourier results.