A Deep Image Prior network trained with an inertial dynamics system converges to a zero-loss solution with an accelerated exponential rate in continuous time, and a discretized version achieves linear convergence with comparable recovery bounds.
, author Montanari, A
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
-
Implicit Regularization of the Deep Inverse Prior Trained with Inertia
A Deep Image Prior network trained with an inertial dynamics system converges to a zero-loss solution with an accelerated exponential rate in continuous time, and a discretized version achieves linear convergence with comparable recovery bounds.