FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3210128071> ?p ?o ?g. }

• W3210128071 abstract "A growing body of research in continual learning is devoted to overcoming the Catastrophic Forgetting of neural networks by designing new algorithms that are more robust to the distribution shifts. While the recent progress in continual learning literature is encouraging, our understanding of what properties of neural networks contribute to catastrophic forgetting is still limited. To address this, instead of focusing on continual learning algorithms, in this work, we focus on the model itself and study the impact of of the neural network architecture on catastrophic forgetting, and show that width has a surprisingly significant effect on forgetting. To explain this effect, we study the learning dynamics of the network from various perspectives such as gradient norm and sparsity, orthogonalization, and lazy training regime. We provide potential explanations that are consistent with the empirical results across different architectures and continual learning benchmarks." @default.
• W3210128071 created "2021-11-08" @default.
• W3210128071 creator A5010129552 @default.
• W3210128071 creator A5038288567 @default.
• W3210128071 creator A5043910056 @default.
• W3210128071 creator A5050499655 @default.
• W3210128071 creator A5070231479 @default.
• W3210128071 creator A5081258844 @default.
• W3210128071 date "2021-10-21" @default.
• W3210128071 modified "2023-09-27" @default.
• W3210128071 title "Wide Neural Networks Forget Less Catastrophically" @default.
• W3210128071 cites W1535804263 @default.
• W3210128071 cites W1682403713 @default.
• W3210128071 cites W1815076433 @default.
• W3210128071 cites W2047057213 @default.
• W3210128071 cites W2107878631 @default.
• W3210128071 cites W2113839990 @default.
• W3210128071 cites W2163605009 @default.
• W3210128071 cites W2183341477 @default.
• W3210128071 cites W2194775991 @default.
• W3210128071 cites W2554616628 @default.
• W3210128071 cites W2557283755 @default.
• W3210128071 cites W2560647685 @default.
• W3210128071 cites W2583761661 @default.
• W3210128071 cites W2737492962 @default.
• W3210128071 cites W2765101016 @default.
• W3210128071 cites W2788388592 @default.
• W3210128071 cites W2806984819 @default.
• W3210128071 cites W2809090039 @default.
• W3210128071 cites W2811024793 @default.
• W3210128071 cites W2899063268 @default.
• W3210128071 cites W2902456977 @default.
• W3210128071 cites W2902625698 @default.
• W3210128071 cites W2903996579 @default.
• W3210128071 cites W2912515466 @default.
• W3210128071 cites W2932484476 @default.
• W3210128071 cites W2947461406 @default.
• W3210128071 cites W2952204734 @default.
• W3210128071 cites W2962698540 @default.
• W3210128071 cites W2962724315 @default.
• W3210128071 cites W2963390791 @default.
• W3210128071 cites W2963393838 @default.
• W3210128071 cites W2963444224 @default.
• W3210128071 cites W2963559848 @default.
• W3210128071 cites W2963588172 @default.
• W3210128071 cites W2964137095 @default.
• W3210128071 cites W2964189064 @default.
• W3210128071 cites W2964790801 @default.
• W3210128071 cites W2970330753 @default.
• W3210128071 cites W2970505118 @default.
• W3210128071 cites W2971043187 @default.
• W3210128071 cites W2981407587 @default.
• W3210128071 cites W3000677656 @default.
• W3210128071 cites W3001279689 @default.
• W3210128071 cites W3008449794 @default.
• W3210128071 cites W3030163527 @default.
• W3210128071 cites W3035525746 @default.
• W3210128071 cites W3037853434 @default.
• W3210128071 cites W3037967334 @default.
• W3210128071 cites W3040607201 @default.
• W3210128071 cites W3098195673 @default.
• W3210128071 cites W3122207284 @default.
• W3210128071 cites W3126553126 @default.
• W3210128071 cites W3132646476 @default.
• W3210128071 cites W3157424867 @default.
• W3210128071 cites W3171460770 @default.
• W3210128071 cites W2426267443 @default.
• W3210128071 hasPublicationYear "2021" @default.
• W3210128071 type Work @default.
• W3210128071 sameAs 3210128071 @default.
• W3210128071 citedByCount "0" @default.
• W3210128071 crossrefType "posted-content" @default.
• W3210128071 hasAuthorship W3210128071A5010129552 @default.
• W3210128071 hasAuthorship W3210128071A5038288567 @default.
• W3210128071 hasAuthorship W3210128071A5043910056 @default.
• W3210128071 hasAuthorship W3210128071A5050499655 @default.
• W3210128071 hasAuthorship W3210128071A5070231479 @default.
• W3210128071 hasAuthorship W3210128071A5081258844 @default.
• W3210128071 hasConcept C108583219 @default.
• W3210128071 hasConcept C11413529 @default.
• W3210128071 hasConcept C119857082 @default.
• W3210128071 hasConcept C120665830 @default.
• W3210128071 hasConcept C121332964 @default.
• W3210128071 hasConcept C154945302 @default.
• W3210128071 hasConcept C15744967 @default.
• W3210128071 hasConcept C180747234 @default.
• W3210128071 hasConcept C192209626 @default.
• W3210128071 hasConcept C41008148 @default.
• W3210128071 hasConcept C47559304 @default.
• W3210128071 hasConcept C50644808 @default.
• W3210128071 hasConcept C7149132 @default.
• W3210128071 hasConceptScore W3210128071C108583219 @default.
• W3210128071 hasConceptScore W3210128071C11413529 @default.
• W3210128071 hasConceptScore W3210128071C119857082 @default.
• W3210128071 hasConceptScore W3210128071C120665830 @default.
• W3210128071 hasConceptScore W3210128071C121332964 @default.
• W3210128071 hasConceptScore W3210128071C154945302 @default.
• W3210128071 hasConceptScore W3210128071C15744967 @default.
• W3210128071 hasConceptScore W3210128071C180747234 @default.
• W3210128071 hasConceptScore W3210128071C192209626 @default.

• next