Matches in SemOpenAlex for { <https://semopenalex.org/work/W4288334269> ?p ?o ?g. }
Showing items 1 to 41 of
41
with 100 items per page.
- W4288334269 abstract "Unrolled neural networks emerged recently as an effective model for learning inverse maps appearing in image restoration tasks. However, their generalization risk (i.e., test mean-squared-error) and its link to network design and train sample size remains mysterious. Leveraging the Stein's Unbiased Risk Estimator (SURE), this paper analyzes the generalization risk with its bias and variance components for recurrent unrolled networks. We particularly investigate the degrees-of-freedom (DOF) component of SURE, trace of the end-to-end network Jacobian, to quantify the prediction variance. We prove that DOF is well-approximated by the weighted textit{path sparsity} of the network under incoherence conditions on the trained weights. Empirically, we examine the SURE components as a function of train sample size for both recurrent and non-recurrent (with many more parameters) unrolled networks. Our key observations indicate that: 1) DOF increases with train sample size and converges to the generalization risk for both recurrent and non-recurrent schemes; 2) recurrent network converges significantly faster (with less train samples) compared with non-recurrent scheme, hence recurrence serves as a regularization for low sample size regimes." @default.
- W4288334269 created "2022-07-28" @default.
- W4288334269 creator A5008348052 @default.
- W4288334269 creator A5018100130 @default.
- W4288334269 creator A5046973419 @default.
- W4288334269 creator A5060755739 @default.
- W4288334269 creator A5065355421 @default.
- W4288334269 creator A5070743328 @default.
- W4288334269 date "2019-06-09" @default.
- W4288334269 modified "2023-09-27" @default.
- W4288334269 title "Degrees of Freedom Analysis of Unrolled Neural Networks" @default.
- W4288334269 doi "https://doi.org/10.48550/arxiv.1906.03742" @default.
- W4288334269 hasPublicationYear "2019" @default.
- W4288334269 type Work @default.
- W4288334269 citedByCount "0" @default.
- W4288334269 crossrefType "posted-content" @default.
- W4288334269 hasAuthorship W4288334269A5008348052 @default.
- W4288334269 hasAuthorship W4288334269A5018100130 @default.
- W4288334269 hasAuthorship W4288334269A5046973419 @default.
- W4288334269 hasAuthorship W4288334269A5060755739 @default.
- W4288334269 hasAuthorship W4288334269A5065355421 @default.
- W4288334269 hasAuthorship W4288334269A5070743328 @default.
- W4288334269 hasBestOaLocation W42883342691 @default.
- W4288334269 hasConcept C121332964 @default.
- W4288334269 hasConcept C154945302 @default.
- W4288334269 hasConcept C208081375 @default.
- W4288334269 hasConcept C41008148 @default.
- W4288334269 hasConcept C50644808 @default.
- W4288334269 hasConcept C62520636 @default.
- W4288334269 hasConceptScore W4288334269C121332964 @default.
- W4288334269 hasConceptScore W4288334269C154945302 @default.
- W4288334269 hasConceptScore W4288334269C208081375 @default.
- W4288334269 hasConceptScore W4288334269C41008148 @default.
- W4288334269 hasConceptScore W4288334269C50644808 @default.
- W4288334269 hasConceptScore W4288334269C62520636 @default.
- W4288334269 hasLocation W42883342691 @default.
- W4288334269 hasOpenAccess W4288334269 @default.
- W4288334269 hasPrimaryLocation W42883342691 @default.
- W4288334269 isParatext "false" @default.
- W4288334269 isRetracted "false" @default.
- W4288334269 workType "article" @default.