FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3207899193> ?p ?o ?g. }

• W3207899193 abstract "Understanding the fundamental mechanism behind the success of deep neural networks is one of the key challenges in the modern machine learning literature. Despite numerous attempts, a solid theoretical analysis is yet to be developed. In this paper, we develop a novel unified framework to reveal a hidden regularization mechanism through the lens of convex optimization. We first show that the training of multiple three-layer ReLU sub-networks with weight decay regularization can be equivalently cast as a convex optimization problem in a higher dimensional space, where sparsity is enforced via a group $ell_1$-norm regularization. Consequently, ReLU networks can be interpreted as high dimensional feature selection methods. More importantly, we then prove that the equivalent convex problem can be globally optimized by a standard convex optimization solver with a polynomial-time complexity with respect to the number of samples and data dimension when the width of the network is fixed. Finally, we numerically validate our theoretical results via experiments involving both synthetic and real datasets." @default.
• W3207899193 created "2021-10-25" @default.
• W3207899193 creator A5001436196 @default.
• W3207899193 creator A5040173784 @default.
• W3207899193 date "2021-10-11" @default.
• W3207899193 modified "2023-09-27" @default.
• W3207899193 title "Global Optimality Beyond Two Layers: Training Deep ReLU Networks via Convex Programs" @default.
• W3207899193 cites W1509722456 @default.
• W3207899193 cites W167786791 @default.
• W3207899193 cites W171072194 @default.
• W3207899193 cites W1985077192 @default.
• W3207899193 cites W2025762001 @default.
• W3207899193 cites W2108383531 @default.
• W3207899193 cites W2128865943 @default.
• W3207899193 cites W2151175710 @default.
• W3207899193 cites W2161278885 @default.
• W3207899193 cites W2181056828 @default.
• W3207899193 cites W2194775991 @default.
• W3207899193 cites W2279098554 @default.
• W3207899193 cites W2401231614 @default.
• W3207899193 cites W2531409750 @default.
• W3207899193 cites W2549139847 @default.
• W3207899193 cites W2608666998 @default.
• W3207899193 cites W2623127191 @default.
• W3207899193 cites W2746420172 @default.
• W3207899193 cites W2750384547 @default.
• W3207899193 cites W2766371994 @default.
• W3207899193 cites W2807299122 @default.
• W3207899193 cites W2920945696 @default.
• W3207899193 cites W2952318479 @default.
• W3207899193 cites W2962930448 @default.
• W3207899193 cites W2963100491 @default.
• W3207899193 cites W2963376662 @default.
• W3207899193 cites W2963383839 @default.
• W3207899193 cites W2963410064 @default.
• W3207899193 cites W2963534251 @default.
• W3207899193 cites W2963687412 @default.
• W3207899193 cites W2963695615 @default.
• W3207899193 cites W2963966702 @default.
• W3207899193 cites W2964350391 @default.
• W3207899193 cites W2964624822 @default.
• W3207899193 cites W2992146231 @default.
• W3207899193 cites W3006926186 @default.
• W3207899193 cites W3007103516 @default.
• W3207899193 cites W3034552778 @default.
• W3207899193 cites W3132829052 @default.
• W3207899193 cites W3135094205 @default.
• W3207899193 cites W3160828921 @default.
• W3207899193 cites W3169249141 @default.
• W3207899193 cites W2290452516 @default.
• W3207899193 hasPublicationYear "2021" @default.
• W3207899193 type Work @default.
• W3207899193 sameAs 3207899193 @default.
• W3207899193 citedByCount "0" @default.
• W3207899193 crossrefType "posted-content" @default.
• W3207899193 hasAuthorship W3207899193A5001436196 @default.
• W3207899193 hasAuthorship W3207899193A5040173784 @default.
• W3207899193 hasConcept C108583219 @default.
• W3207899193 hasConcept C112680207 @default.
• W3207899193 hasConcept C11413529 @default.
• W3207899193 hasConcept C12108790 @default.
• W3207899193 hasConcept C126255220 @default.
• W3207899193 hasConcept C137836250 @default.
• W3207899193 hasConcept C154945302 @default.
• W3207899193 hasConcept C157972887 @default.
• W3207899193 hasConcept C17744445 @default.
• W3207899193 hasConcept C191795146 @default.
• W3207899193 hasConcept C199539241 @default.
• W3207899193 hasConcept C202444582 @default.
• W3207899193 hasConcept C2524010 @default.
• W3207899193 hasConcept C2776135515 @default.
• W3207899193 hasConcept C2984842247 @default.
• W3207899193 hasConcept C3019722297 @default.
• W3207899193 hasConcept C33676613 @default.
• W3207899193 hasConcept C33923547 @default.
• W3207899193 hasConcept C41008148 @default.
• W3207899193 hasConcept C50644808 @default.
• W3207899193 hasConcept C79248915 @default.
• W3207899193 hasConceptScore W3207899193C108583219 @default.
• W3207899193 hasConceptScore W3207899193C112680207 @default.
• W3207899193 hasConceptScore W3207899193C11413529 @default.
• W3207899193 hasConceptScore W3207899193C12108790 @default.
• W3207899193 hasConceptScore W3207899193C126255220 @default.
• W3207899193 hasConceptScore W3207899193C137836250 @default.
• W3207899193 hasConceptScore W3207899193C154945302 @default.
• W3207899193 hasConceptScore W3207899193C157972887 @default.
• W3207899193 hasConceptScore W3207899193C17744445 @default.
• W3207899193 hasConceptScore W3207899193C191795146 @default.
• W3207899193 hasConceptScore W3207899193C199539241 @default.
• W3207899193 hasConceptScore W3207899193C202444582 @default.
• W3207899193 hasConceptScore W3207899193C2524010 @default.
• W3207899193 hasConceptScore W3207899193C2776135515 @default.
• W3207899193 hasConceptScore W3207899193C2984842247 @default.
• W3207899193 hasConceptScore W3207899193C3019722297 @default.
• W3207899193 hasConceptScore W3207899193C33676613 @default.
• W3207899193 hasConceptScore W3207899193C33923547 @default.
• W3207899193 hasConceptScore W3207899193C41008148 @default.
• W3207899193 hasConceptScore W3207899193C50644808 @default.
• W3207899193 hasConceptScore W3207899193C79248915 @default.
• W3207899193 hasLocation W32078991931 @default.

• next