Matches in SemOpenAlex for { <https://semopenalex.org/work/W3170349722> ?p ?o ?g. }
- W3170349722 endingPage "2192" @default.
- W3170349722 startingPage "2163" @default.
- W3170349722 abstract "Deep learning is often criticized by two serious issues that rarely exist in natural nervous systems: overfitting and catastrophic forgetting. It can even memorize randomly labeled data, which has little knowledge behind the instance-label pairs. When a deep network continually learns over time by accommodating new tasks, it usually quickly overwrites the knowledge learned from previous tasks. Referred to as the neural variability, it is well known in neuroscience that human brain reactions exhibit substantial variability even in response to the same stimulus. This mechanism balances accuracy and plasticity/flexibility in the motor learning of natural nervous systems. Thus, it motivates us to design a similar mechanism, named artificial neural variability (ANV), that helps artificial neural networks learn some advantages from natural neural networks. We rigorously prove that ANV plays as an implicit regularizer of the mutual information between the training data and the learned model. This result theoretically guarantees ANV a strictly improved generalizability, robustness to label noise, and robustness to catastrophic forgetting. We then devise a neural variable risk minimization (NVRM) framework and neural variable optimizers to achieve ANV for conventional network architectures in practice. The empirical studies demonstrate that NVRM can effectively relieve overfitting, label noise memorization, and catastrophic forgetting at negligible costs." @default.
- W3170349722 created "2021-06-22" @default.
- W3170349722 creator A5001819736 @default.
- W3170349722 creator A5017703463 @default.
- W3170349722 creator A5032483871 @default.
- W3170349722 creator A5060421432 @default.
- W3170349722 creator A5066773635 @default.
- W3170349722 creator A5072744508 @default.
- W3170349722 date "2021-07-26" @default.
- W3170349722 modified "2023-09-30" @default.
- W3170349722 title "Artificial Neural Variability for Deep Learning: On Overfitting, Noise Memorization, and Catastrophic Forgetting" @default.
- W3170349722 cites W1682403713 @default.
- W3170349722 cites W1864407449 @default.
- W3170349722 cites W1971735090 @default.
- W3170349722 cites W1984367183 @default.
- W3170349722 cites W1988245436 @default.
- W3170349722 cites W1998794714 @default.
- W3170349722 cites W2004132309 @default.
- W3170349722 cites W2029029543 @default.
- W3170349722 cites W2040499723 @default.
- W3170349722 cites W2041540302 @default.
- W3170349722 cites W2047229728 @default.
- W3170349722 cites W2103496339 @default.
- W3170349722 cites W2105380935 @default.
- W3170349722 cites W2124136621 @default.
- W3170349722 cites W2124227140 @default.
- W3170349722 cites W2137983211 @default.
- W3170349722 cites W2149479912 @default.
- W3170349722 cites W2171776652 @default.
- W3170349722 cites W2257979135 @default.
- W3170349722 cites W2560647685 @default.
- W3170349722 cites W2592929672 @default.
- W3170349722 cites W2788388592 @default.
- W3170349722 cites W2890638274 @default.
- W3170349722 cites W2900832763 @default.
- W3170349722 cites W2912811302 @default.
- W3170349722 cites W2919115771 @default.
- W3170349722 cites W2938101602 @default.
- W3170349722 cites W2962720772 @default.
- W3170349722 cites W4244569433 @default.
- W3170349722 doi "https://doi.org/10.1162/neco_a_01403" @default.
- W3170349722 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/34310675" @default.
- W3170349722 hasPublicationYear "2021" @default.
- W3170349722 type Work @default.
- W3170349722 sameAs 3170349722 @default.
- W3170349722 citedByCount "22" @default.
- W3170349722 countsByYear W31703497222020 @default.
- W3170349722 countsByYear W31703497222021 @default.
- W3170349722 countsByYear W31703497222022 @default.
- W3170349722 countsByYear W31703497222023 @default.
- W3170349722 crossrefType "journal-article" @default.
- W3170349722 hasAuthorship W3170349722A5001819736 @default.
- W3170349722 hasAuthorship W3170349722A5017703463 @default.
- W3170349722 hasAuthorship W3170349722A5032483871 @default.
- W3170349722 hasAuthorship W3170349722A5060421432 @default.
- W3170349722 hasAuthorship W3170349722A5066773635 @default.
- W3170349722 hasAuthorship W3170349722A5072744508 @default.
- W3170349722 hasBestOaLocation W31703497222 @default.
- W3170349722 hasConcept C104317684 @default.
- W3170349722 hasConcept C115961682 @default.
- W3170349722 hasConcept C119857082 @default.
- W3170349722 hasConcept C154945302 @default.
- W3170349722 hasConcept C15744967 @default.
- W3170349722 hasConcept C180747234 @default.
- W3170349722 hasConcept C185592680 @default.
- W3170349722 hasConcept C22019652 @default.
- W3170349722 hasConcept C2776145597 @default.
- W3170349722 hasConcept C2984842247 @default.
- W3170349722 hasConcept C30038468 @default.
- W3170349722 hasConcept C41008148 @default.
- W3170349722 hasConcept C50644808 @default.
- W3170349722 hasConcept C55493867 @default.
- W3170349722 hasConcept C63479239 @default.
- W3170349722 hasConcept C7149132 @default.
- W3170349722 hasConcept C99498987 @default.
- W3170349722 hasConceptScore W3170349722C104317684 @default.
- W3170349722 hasConceptScore W3170349722C115961682 @default.
- W3170349722 hasConceptScore W3170349722C119857082 @default.
- W3170349722 hasConceptScore W3170349722C154945302 @default.
- W3170349722 hasConceptScore W3170349722C15744967 @default.
- W3170349722 hasConceptScore W3170349722C180747234 @default.
- W3170349722 hasConceptScore W3170349722C185592680 @default.
- W3170349722 hasConceptScore W3170349722C22019652 @default.
- W3170349722 hasConceptScore W3170349722C2776145597 @default.
- W3170349722 hasConceptScore W3170349722C2984842247 @default.
- W3170349722 hasConceptScore W3170349722C30038468 @default.
- W3170349722 hasConceptScore W3170349722C41008148 @default.
- W3170349722 hasConceptScore W3170349722C50644808 @default.
- W3170349722 hasConceptScore W3170349722C55493867 @default.
- W3170349722 hasConceptScore W3170349722C63479239 @default.
- W3170349722 hasConceptScore W3170349722C7149132 @default.
- W3170349722 hasConceptScore W3170349722C99498987 @default.
- W3170349722 hasIssue "8" @default.
- W3170349722 hasLocation W31703497221 @default.
- W3170349722 hasLocation W31703497222 @default.
- W3170349722 hasLocation W31703497223 @default.
- W3170349722 hasOpenAccess W3170349722 @default.
- W3170349722 hasPrimaryLocation W31703497221 @default.