Matches in SemOpenAlex for { <https://semopenalex.org/work/W3136604563> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3136604563 abstract "Solving partial differential equations (PDEs) by parametrizing its solution by neural networks (NNs) has been popular in the past a few years. However, different types of loss functions can be proposed for the same PDE. For the Poisson equation, the loss function can be based on the weak formulation of energy variation or the least squares method, which leads to the deep Ritz model and deep Galerkin model, respectively. But loss landscapes from these different models give arise to different practical performance of training the NN parameters. To investigate and understand such practical differences, we propose to compare the loss landscapes of these models, which are both high dimensional and highly non-convex. In such settings, the is more important than the traditional eigenvalue analysis to describe the non-convexity. We contribute to the landscape comparisons by proposing a index to scientifically and quantitatively describe the heuristic concept of roughness of landscape around minimizers. This index is based on random projections and the variance of (normalized) total variation for one dimensional projected functions, and it is efficient to compute. A large index hints an oscillatory landscape profile as a severe challenge for the first order optimization method. We apply this index to the two models for the Poisson equation and our empirical results reveal a consistent general observation that the landscapes from the deep Galerkin method around its local minimizers are less rough than the deep Ritz method, which supports the observed gain in accuracy of the deep Galerkin method." @default.
- W3136604563 created "2021-03-29" @default.
- W3136604563 creator A5000360043 @default.
- W3136604563 creator A5025195102 @default.
- W3136604563 creator A5047988362 @default.
- W3136604563 creator A5068296887 @default.
- W3136604563 date "2021-03-20" @default.
- W3136604563 modified "2023-09-27" @default.
- W3136604563 title "Understanding Loss Landscapes of Neural Network Models in Solving Partial Differential Equations." @default.
- W3136604563 cites W1547249381 @default.
- W3136604563 cites W1855169650 @default.
- W3136604563 cites W2523060838 @default.
- W3136604563 cites W2586189899 @default.
- W3136604563 cites W2731468224 @default.
- W3136604563 cites W2760972773 @default.
- W3136604563 cites W2892675615 @default.
- W3136604563 cites W2911183980 @default.
- W3136604563 cites W2918085900 @default.
- W3136604563 cites W2949650786 @default.
- W3136604563 cites W2962781506 @default.
- W3136604563 cites W2962933129 @default.
- W3136604563 cites W2963317585 @default.
- W3136604563 cites W2963794891 @default.
- W3136604563 cites W2964072432 @default.
- W3136604563 cites W2964121744 @default.
- W3136604563 cites W2964160102 @default.
- W3136604563 cites W2993505246 @default.
- W3136604563 cites W3033623076 @default.
- W3136604563 cites W3047103993 @default.
- W3136604563 cites W3082030075 @default.
- W3136604563 cites W3115761617 @default.
- W3136604563 cites W3137695714 @default.
- W3136604563 hasPublicationYear "2021" @default.
- W3136604563 type Work @default.
- W3136604563 sameAs 3136604563 @default.
- W3136604563 citedByCount "0" @default.
- W3136604563 crossrefType "posted-content" @default.
- W3136604563 hasAuthorship W3136604563A5000360043 @default.
- W3136604563 hasAuthorship W3136604563A5025195102 @default.
- W3136604563 hasAuthorship W3136604563A5047988362 @default.
- W3136604563 hasAuthorship W3136604563A5068296887 @default.
- W3136604563 hasConcept C100906024 @default.
- W3136604563 hasConcept C105795698 @default.
- W3136604563 hasConcept C121332964 @default.
- W3136604563 hasConcept C126255220 @default.
- W3136604563 hasConcept C134306372 @default.
- W3136604563 hasConcept C135628077 @default.
- W3136604563 hasConcept C154945302 @default.
- W3136604563 hasConcept C173801870 @default.
- W3136604563 hasConcept C186899397 @default.
- W3136604563 hasConcept C28826006 @default.
- W3136604563 hasConcept C33923547 @default.
- W3136604563 hasConcept C41008148 @default.
- W3136604563 hasConcept C50644808 @default.
- W3136604563 hasConcept C93779851 @default.
- W3136604563 hasConcept C96716743 @default.
- W3136604563 hasConcept C97355855 @default.
- W3136604563 hasConceptScore W3136604563C100906024 @default.
- W3136604563 hasConceptScore W3136604563C105795698 @default.
- W3136604563 hasConceptScore W3136604563C121332964 @default.
- W3136604563 hasConceptScore W3136604563C126255220 @default.
- W3136604563 hasConceptScore W3136604563C134306372 @default.
- W3136604563 hasConceptScore W3136604563C135628077 @default.
- W3136604563 hasConceptScore W3136604563C154945302 @default.
- W3136604563 hasConceptScore W3136604563C173801870 @default.
- W3136604563 hasConceptScore W3136604563C186899397 @default.
- W3136604563 hasConceptScore W3136604563C28826006 @default.
- W3136604563 hasConceptScore W3136604563C33923547 @default.
- W3136604563 hasConceptScore W3136604563C41008148 @default.
- W3136604563 hasConceptScore W3136604563C50644808 @default.
- W3136604563 hasConceptScore W3136604563C93779851 @default.
- W3136604563 hasConceptScore W3136604563C96716743 @default.
- W3136604563 hasConceptScore W3136604563C97355855 @default.
- W3136604563 hasLocation W31366045631 @default.
- W3136604563 hasOpenAccess W3136604563 @default.
- W3136604563 hasPrimaryLocation W31366045631 @default.
- W3136604563 hasRelatedWork W1884988597 @default.
- W3136604563 hasRelatedWork W1946068080 @default.
- W3136604563 hasRelatedWork W1992077634 @default.
- W3136604563 hasRelatedWork W2169704132 @default.
- W3136604563 hasRelatedWork W2307551484 @default.
- W3136604563 hasRelatedWork W2467245180 @default.
- W3136604563 hasRelatedWork W2734265911 @default.
- W3136604563 hasRelatedWork W2793980235 @default.
- W3136604563 hasRelatedWork W2890473586 @default.
- W3136604563 hasRelatedWork W2896469289 @default.
- W3136604563 hasRelatedWork W2905630886 @default.
- W3136604563 hasRelatedWork W2938255066 @default.
- W3136604563 hasRelatedWork W2947511419 @default.
- W3136604563 hasRelatedWork W2952034592 @default.
- W3136604563 hasRelatedWork W2953988481 @default.
- W3136604563 hasRelatedWork W2970217468 @default.
- W3136604563 hasRelatedWork W3042743256 @default.
- W3136604563 hasRelatedWork W3106072647 @default.
- W3136604563 hasRelatedWork W3209539220 @default.
- W3136604563 hasRelatedWork W3211861249 @default.
- W3136604563 isParatext "false" @default.
- W3136604563 isRetracted "false" @default.
- W3136604563 magId "3136604563" @default.
- W3136604563 workType "article" @default.