Low-tubal-rank tensors can be recovered from independent compressed measurements of each lateral slice using alternating minimization with spectral initialization, and a preconditioned variant achieves condition-number-independent iteration complexity.
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
-
Non-Convex Tensor Recovery from Local Measurements
Low-tubal-rank tensors can be recovered from independent compressed measurements of each lateral slice using alternating minimization with spectral initialization, and a preconditioned variant achieves condition-number-independent iteration complexity.