FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3213505436> ?p ?o ?g. }

• W3213505436 abstract "Recent research shows that the dynamics of an infinitely wide neural network (NN) trained by gradient descent can be characterized by Neural Tangent Kernel (NTK) citep{jacot2018neural}. Under the squared loss, the infinite-width NN trained by gradient descent with an infinitely small learning rate is equivalent to kernel regression with NTK citep{arora2019exact}. However, the equivalence is only known for ridge regression currently citep{arora2019harnessing}, while the equivalence between NN and other kernel machines (KMs), e.g. support vector machine (SVM), remains unknown. Therefore, in this work, we propose to establish the equivalence between NN and SVM, and specifically, the infinitely wide NN trained by soft margin loss and the standard soft margin SVM with NTK trained by subgradient descent. Our main theoretical results include establishing the equivalences between NNs and a broad family of $ell_2$ regularized KMs with finite-width bounds, which cannot be handled by prior work, and showing that every finite-width NN trained by such regularized loss functions is approximately a KM. Furthermore, we demonstrate our theory can enable three practical applications, including (i) textit{non-vacuous} generalization bound of NN via the corresponding KM; (ii) textit{non-trivial} robustness certificate for the infinite-width NN (while existing robustness verification methods would provide vacuous bounds); (iii) intrinsically more robust infinite-width NNs than those from previous kernel regression. Our code for the experiments is available at url{https://github.com/leslie-CH/equiv-nn-svm}." @default.
• W3213505436 created "2021-11-22" @default.
• W3213505436 creator A5000385187 @default.
• W3213505436 creator A5007306953 @default.
• W3213505436 creator A5037975686 @default.
• W3213505436 creator A5049043092 @default.
• W3213505436 date "2021-11-11" @default.
• W3213505436 modified "2023-09-23" @default.
• W3213505436 title "On the Equivalence between Neural Network and Support Vector Machine" @default.
• W3213505436 cites W1409984952 @default.
• W3213505436 cites W1510073064 @default.
• W3213505436 cites W1522579744 @default.
• W3213505436 cites W1529331786 @default.
• W3213505436 cites W1546411676 @default.
• W3213505436 cites W1560724230 @default.
• W3213505436 cites W1563088657 @default.
• W3213505436 cites W1698155719 @default.
• W3213505436 cites W1944672 @default.
• W3213505436 cites W2007154098 @default.
• W3213505436 cites W2043701535 @default.
• W3213505436 cites W2087347434 @default.
• W3213505436 cites W2112796928 @default.
• W3213505436 cites W2119821739 @default.
• W3213505436 cites W2125993116 @default.
• W3213505436 cites W2143956139 @default.
• W3213505436 cites W2165966284 @default.
• W3213505436 cites W2167608136 @default.
• W3213505436 cites W2292929658 @default.
• W3213505436 cites W2529714286 @default.
• W3213505436 cites W2566079294 @default.
• W3213505436 cites W2579923771 @default.
• W3213505436 cites W2604117713 @default.
• W3213505436 cites W2766557427 @default.
• W3213505436 cites W2766678531 @default.
• W3213505436 cites W2785626633 @default.
• W3213505436 cites W2800415562 @default.
• W3213505436 cites W2809090039 @default.
• W3213505436 cites W2894604724 @default.
• W3213505436 cites W2898963688 @default.
• W3213505436 cites W2899748887 @default.
• W3213505436 cites W2911867426 @default.
• W3213505436 cites W2946840143 @default.
• W3213505436 cites W2946985165 @default.
• W3213505436 cites W2962698540 @default.
• W3213505436 cites W2963424284 @default.
• W3213505436 cites W2963440492 @default.
• W3213505436 cites W2963656735 @default.
• W3213505436 cites W2970032917 @default.
• W3213505436 cites W2970971581 @default.
• W3213505436 cites W2971043187 @default.
• W3213505436 cites W2977953007 @default.
• W3213505436 cites W2994747787 @default.
• W3213505436 cites W2994872659 @default.
• W3213505436 cites W3004459530 @default.
• W3213505436 cites W3015989168 @default.
• W3213505436 cites W3039141380 @default.
• W3213505436 cites W3102429474 @default.
• W3213505436 cites W3105547603 @default.
• W3213505436 cites W3108435811 @default.
• W3213505436 cites W3158239976 @default.
• W3213505436 cites W3165421101 @default.
• W3213505436 cites W3181384422 @default.
• W3213505436 cites W607505555 @default.
• W3213505436 doi "https://doi.org/10.48550/arxiv.2111.06063" @default.
• W3213505436 hasPublicationYear "2021" @default.
• W3213505436 type Work @default.
• W3213505436 sameAs 3213505436 @default.
• W3213505436 citedByCount "0" @default.
• W3213505436 crossrefType "posted-content" @default.
• W3213505436 hasAuthorship W3213505436A5000385187 @default.
• W3213505436 hasAuthorship W3213505436A5007306953 @default.
• W3213505436 hasAuthorship W3213505436A5037975686 @default.
• W3213505436 hasAuthorship W3213505436A5049043092 @default.
• W3213505436 hasBestOaLocation W32135054361 @default.
• W3213505436 hasConcept C104317684 @default.
• W3213505436 hasConcept C11413529 @default.
• W3213505436 hasConcept C118615104 @default.
• W3213505436 hasConcept C119857082 @default.
• W3213505436 hasConcept C12267149 @default.
• W3213505436 hasConcept C153258448 @default.
• W3213505436 hasConcept C154945302 @default.
• W3213505436 hasConcept C185592680 @default.
• W3213505436 hasConcept C202444582 @default.
• W3213505436 hasConcept C22324862 @default.
• W3213505436 hasConcept C2780069185 @default.
• W3213505436 hasConcept C28826006 @default.
• W3213505436 hasConcept C33923547 @default.
• W3213505436 hasConcept C41008148 @default.
• W3213505436 hasConcept C50644808 @default.
• W3213505436 hasConcept C55493867 @default.
• W3213505436 hasConcept C63479239 @default.
• W3213505436 hasConcept C74193536 @default.
• W3213505436 hasConcept C774472 @default.
• W3213505436 hasConceptScore W3213505436C104317684 @default.
• W3213505436 hasConceptScore W3213505436C11413529 @default.
• W3213505436 hasConceptScore W3213505436C118615104 @default.
• W3213505436 hasConceptScore W3213505436C119857082 @default.
• W3213505436 hasConceptScore W3213505436C12267149 @default.
• W3213505436 hasConceptScore W3213505436C153258448 @default.
• W3213505436 hasConceptScore W3213505436C154945302 @default.

• next