Matches in SemOpenAlex for { <https://semopenalex.org/work/W2891857315> ?p ?o ?g. }
Showing items 1 to 90 of
90
with 100 items per page.
- W2891857315 abstract "We study the dynamics of gradient descent on objective functions of the form $f(prod_{i=1}^{k} w_i)$ (with respect to scalar parameters $w_1,ldots,w_k$), which arise in the context of training depth-$k$ linear neural networks. We prove that for standard random initializations, and under mild assumptions on $f$, the number of iterations required for convergence scales exponentially with the depth $k$. We also show empirically that this phenomenon can occur in higher dimensions, where each $w_i$ is a matrix. This highlights a potential obstacle in understanding the convergence of gradient-based methods for deep linear neural networks, where $k$ is large." @default.
- W2891857315 created "2018-09-27" @default.
- W2891857315 creator A5017753135 @default.
- W2891857315 date "2018-09-23" @default.
- W2891857315 modified "2023-10-17" @default.
- W2891857315 title "Exponential Convergence Time of Gradient Descent for One-Dimensional Deep Linear Neural Networks" @default.
- W2891857315 cites W1278605765 @default.
- W2891857315 cites W1533861849 @default.
- W2891857315 cites W2125930537 @default.
- W2891857315 cites W2593380010 @default.
- W2891857315 cites W2788800397 @default.
- W2891857315 cites W2886685759 @default.
- W2891857315 cites W2894972989 @default.
- W2891857315 cites W2952574409 @default.
- W2891857315 cites W2963248893 @default.
- W2891857315 cites W2963446085 @default.
- W2891857315 cites W2963570896 @default.
- W2891857315 cites W2963837241 @default.
- W2891857315 cites W2964072429 @default.
- W2891857315 cites W2964220724 @default.
- W2891857315 hasPublicationYear "2018" @default.
- W2891857315 type Work @default.
- W2891857315 sameAs 2891857315 @default.
- W2891857315 citedByCount "2" @default.
- W2891857315 countsByYear W28918573152019 @default.
- W2891857315 crossrefType "posted-content" @default.
- W2891857315 hasAuthorship W2891857315A5017753135 @default.
- W2891857315 hasConcept C127313418 @default.
- W2891857315 hasConcept C134306372 @default.
- W2891857315 hasConcept C151376022 @default.
- W2891857315 hasConcept C151730666 @default.
- W2891857315 hasConcept C153258448 @default.
- W2891857315 hasConcept C154945302 @default.
- W2891857315 hasConcept C162324750 @default.
- W2891857315 hasConcept C206688291 @default.
- W2891857315 hasConcept C2524010 @default.
- W2891857315 hasConcept C2777303404 @default.
- W2891857315 hasConcept C2779343474 @default.
- W2891857315 hasConcept C28826006 @default.
- W2891857315 hasConcept C33923547 @default.
- W2891857315 hasConcept C41008148 @default.
- W2891857315 hasConcept C50522688 @default.
- W2891857315 hasConcept C50644808 @default.
- W2891857315 hasConcept C57691317 @default.
- W2891857315 hasConcept C75235859 @default.
- W2891857315 hasConceptScore W2891857315C127313418 @default.
- W2891857315 hasConceptScore W2891857315C134306372 @default.
- W2891857315 hasConceptScore W2891857315C151376022 @default.
- W2891857315 hasConceptScore W2891857315C151730666 @default.
- W2891857315 hasConceptScore W2891857315C153258448 @default.
- W2891857315 hasConceptScore W2891857315C154945302 @default.
- W2891857315 hasConceptScore W2891857315C162324750 @default.
- W2891857315 hasConceptScore W2891857315C206688291 @default.
- W2891857315 hasConceptScore W2891857315C2524010 @default.
- W2891857315 hasConceptScore W2891857315C2777303404 @default.
- W2891857315 hasConceptScore W2891857315C2779343474 @default.
- W2891857315 hasConceptScore W2891857315C28826006 @default.
- W2891857315 hasConceptScore W2891857315C33923547 @default.
- W2891857315 hasConceptScore W2891857315C41008148 @default.
- W2891857315 hasConceptScore W2891857315C50522688 @default.
- W2891857315 hasConceptScore W2891857315C50644808 @default.
- W2891857315 hasConceptScore W2891857315C57691317 @default.
- W2891857315 hasConceptScore W2891857315C75235859 @default.
- W2891857315 hasLocation W28918573151 @default.
- W2891857315 hasOpenAccess W2891857315 @default.
- W2891857315 hasPrimaryLocation W28918573151 @default.
- W2891857315 hasRelatedWork W2016937404 @default.
- W2891857315 hasRelatedWork W2018549926 @default.
- W2891857315 hasRelatedWork W2044478460 @default.
- W2891857315 hasRelatedWork W2084687242 @default.
- W2891857315 hasRelatedWork W2105875671 @default.
- W2891857315 hasRelatedWork W2389056905 @default.
- W2891857315 hasRelatedWork W2601280495 @default.
- W2891857315 hasRelatedWork W2807729763 @default.
- W2891857315 hasRelatedWork W2810756260 @default.
- W2891857315 hasRelatedWork W2962688058 @default.
- W2891857315 hasRelatedWork W2964344217 @default.
- W2891857315 hasRelatedWork W2966228138 @default.
- W2891857315 hasRelatedWork W2980820424 @default.
- W2891857315 hasRelatedWork W2995553068 @default.
- W2891857315 hasRelatedWork W3008206053 @default.
- W2891857315 hasRelatedWork W3046539760 @default.
- W2891857315 hasRelatedWork W3090423162 @default.
- W2891857315 hasRelatedWork W3126746910 @default.
- W2891857315 hasRelatedWork W3173113055 @default.
- W2891857315 hasRelatedWork W914658793 @default.
- W2891857315 isParatext "false" @default.
- W2891857315 isRetracted "false" @default.
- W2891857315 magId "2891857315" @default.
- W2891857315 workType "article" @default.