Depth-two exponential randomized signatures are claimed to reconstruct up to d^(N+1) signature features from hidden dimension N, based on new linear independence results for tree-like vector fields.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.CA 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Signature Reconstruction from Randomized Signatures
Depth-two exponential randomized signatures are claimed to reconstruct up to d^(N+1) signature features from hidden dimension N, based on new linear independence results for tree-like vector fields.