Matches in SemOpenAlex for { <https://semopenalex.org/work/W3088824160> ?p ?o ?g. }
- W3088824160 abstract "Regularization improves generalization of supervised models to out-of-sample data. Prior works have shown that prediction in the causal direction (effect from cause) results in lower testing error than the anti-causal direction. However, existing regularization methods are agnostic of causality. We introduce Causal Structure Learning (CASTLE) regularization and propose to regularize a neural network by jointly learning the causal relationships between variables. CASTLE learns the causal directed acyclical graph (DAG) as an adjacency matrix embedded in the neural network's input layers, thereby facilitating the discovery of optimal predictors. Furthermore, CASTLE efficiently reconstructs only the features in the causal DAG that have a causal neighbor, whereas reconstruction-based regularizers suboptimally reconstruct all input features. We provide a theoretical generalization bound for our approach and conduct experiments on a plethora of synthetic and real publicly available datasets demonstrating that CASTLE consistently leads to better out-of-sample predictions as compared to other popular benchmark regularizers." @default.
- W3088824160 created "2020-10-01" @default.
- W3088824160 creator A5012339002 @default.
- W3088824160 creator A5077577005 @default.
- W3088824160 creator A5080577523 @default.
- W3088824160 date "2020-09-28" @default.
- W3088824160 modified "2023-09-23" @default.
- W3088824160 title "CASTLE: Regularization via Auxiliary Causal Graph Discovery" @default.
- W3088824160 cites W1815076433 @default.
- W3088824160 cites W1836465849 @default.
- W3088824160 cites W1845812186 @default.
- W3088824160 cites W1904365287 @default.
- W3088824160 cites W2007738694 @default.
- W3088824160 cites W2025768430 @default.
- W3088824160 cites W2047028564 @default.
- W3088824160 cites W2062714872 @default.
- W3088824160 cites W2095705004 @default.
- W3088824160 cites W2106748239 @default.
- W3088824160 cites W2109605413 @default.
- W3088824160 cites W2110798204 @default.
- W3088824160 cites W2116314036 @default.
- W3088824160 cites W2135046866 @default.
- W3088824160 cites W2143891888 @default.
- W3088824160 cites W2144020560 @default.
- W3088824160 cites W2152722485 @default.
- W3088824160 cites W2153929442 @default.
- W3088824160 cites W2162262658 @default.
- W3088824160 cites W2163605009 @default.
- W3088824160 cites W2166469100 @default.
- W3088824160 cites W2181581468 @default.
- W3088824160 cites W2237537322 @default.
- W3088824160 cites W2523060838 @default.
- W3088824160 cites W2552297262 @default.
- W3088824160 cites W2557283755 @default.
- W3088824160 cites W2784407808 @default.
- W3088824160 cites W2790376986 @default.
- W3088824160 cites W2889900103 @default.
- W3088824160 cites W2890732907 @default.
- W3088824160 cites W2890964657 @default.
- W3088824160 cites W2891620446 @default.
- W3088824160 cites W2891765548 @default.
- W3088824160 cites W2945458902 @default.
- W3088824160 cites W2949146054 @default.
- W3088824160 cites W2962695761 @default.
- W3088824160 cites W2963226480 @default.
- W3088824160 cites W2963285844 @default.
- W3088824160 cites W2963334011 @default.
- W3088824160 cites W2963399829 @default.
- W3088824160 cites W2963528347 @default.
- W3088824160 cites W2963663068 @default.
- W3088824160 cites W2970670424 @default.
- W3088824160 cites W2971083842 @default.
- W3088824160 cites W2971310236 @default.
- W3088824160 cites W2975424829 @default.
- W3088824160 cites W2995379565 @default.
- W3088824160 cites W3037834521 @default.
- W3088824160 hasPublicationYear "2020" @default.
- W3088824160 type Work @default.
- W3088824160 sameAs 3088824160 @default.
- W3088824160 citedByCount "3" @default.
- W3088824160 countsByYear W30888241602020 @default.
- W3088824160 countsByYear W30888241602021 @default.
- W3088824160 crossrefType "posted-content" @default.
- W3088824160 hasAuthorship W3088824160A5012339002 @default.
- W3088824160 hasAuthorship W3088824160A5077577005 @default.
- W3088824160 hasAuthorship W3088824160A5080577523 @default.
- W3088824160 hasConcept C105795698 @default.
- W3088824160 hasConcept C110484373 @default.
- W3088824160 hasConcept C11413529 @default.
- W3088824160 hasConcept C11671645 @default.
- W3088824160 hasConcept C119857082 @default.
- W3088824160 hasConcept C121332964 @default.
- W3088824160 hasConcept C132525143 @default.
- W3088824160 hasConcept C13280743 @default.
- W3088824160 hasConcept C134306372 @default.
- W3088824160 hasConcept C154945302 @default.
- W3088824160 hasConcept C163504300 @default.
- W3088824160 hasConcept C177148314 @default.
- W3088824160 hasConcept C180356752 @default.
- W3088824160 hasConcept C185798385 @default.
- W3088824160 hasConcept C205649164 @default.
- W3088824160 hasConcept C2776135515 @default.
- W3088824160 hasConcept C33923547 @default.
- W3088824160 hasConcept C41008148 @default.
- W3088824160 hasConcept C50644808 @default.
- W3088824160 hasConcept C62520636 @default.
- W3088824160 hasConcept C64357122 @default.
- W3088824160 hasConcept C74197172 @default.
- W3088824160 hasConcept C80444323 @default.
- W3088824160 hasConceptScore W3088824160C105795698 @default.
- W3088824160 hasConceptScore W3088824160C110484373 @default.
- W3088824160 hasConceptScore W3088824160C11413529 @default.
- W3088824160 hasConceptScore W3088824160C11671645 @default.
- W3088824160 hasConceptScore W3088824160C119857082 @default.
- W3088824160 hasConceptScore W3088824160C121332964 @default.
- W3088824160 hasConceptScore W3088824160C132525143 @default.
- W3088824160 hasConceptScore W3088824160C13280743 @default.
- W3088824160 hasConceptScore W3088824160C134306372 @default.
- W3088824160 hasConceptScore W3088824160C154945302 @default.
- W3088824160 hasConceptScore W3088824160C163504300 @default.