A deep neural network estimator using a continuous hinge-type surrogate loss achieves minimax-optimal boundary recovery rates up to logs for piecewise smooth boundaries in unlabeled noisy images.
arXiv preprint arXiv:2506.14899 , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ME 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Rate-optimal neural boundary detection from unlabeled noisy images
A deep neural network estimator using a continuous hinge-type surrogate loss achieves minimax-optimal boundary recovery rates up to logs for piecewise smooth boundaries in unlabeled noisy images.