Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287162887> ?p ?o ?g. }
Showing items 1 to 63 of
63
with 100 items per page.
- W4287162887 abstract "Informally, a 'spurious correlation' is the dependence of a model on some aspect of the input data that an analyst thinks shouldn't matter. In machine learning, these have a know-it-when-you-see-it character; e.g., changing the gender of a sentence's subject changes a sentiment predictor's output. To check for spurious correlations, we can 'stress test' models by perturbing irrelevant parts of input data and seeing if model predictions change. In this paper, we study stress testing using the tools of causal inference. We introduce counterfactual invariance as a formalization of the requirement that changing irrelevant parts of the input shouldn't change model predictions. We connect counterfactual invariance to out-of-domain model performance, and provide practical schemes for learning (approximately) counterfactual invariant predictors (without access to counterfactual examples). It turns out that both the means and implications of counterfactual invariance depend fundamentally on the true underlying causal structure of the data -- in particular, whether the label causes the features or the features cause the label. Distinct causal structures require distinct regularization schemes to induce counterfactual invariance. Similarly, counterfactual invariance implies different domain shift guarantees depending on the underlying causal structure. This theory is supported by empirical results on text classification." @default.
- W4287162887 created "2022-07-25" @default.
- W4287162887 creator A5047699861 @default.
- W4287162887 creator A5059293780 @default.
- W4287162887 creator A5060694111 @default.
- W4287162887 creator A5077406767 @default.
- W4287162887 date "2021-05-31" @default.
- W4287162887 modified "2023-09-27" @default.
- W4287162887 title "Counterfactual Invariance to Spurious Correlations: Why and How to Pass Stress Tests" @default.
- W4287162887 doi "https://doi.org/10.48550/arxiv.2106.00545" @default.
- W4287162887 hasPublicationYear "2021" @default.
- W4287162887 type Work @default.
- W4287162887 citedByCount "0" @default.
- W4287162887 crossrefType "posted-content" @default.
- W4287162887 hasAuthorship W4287162887A5047699861 @default.
- W4287162887 hasAuthorship W4287162887A5059293780 @default.
- W4287162887 hasAuthorship W4287162887A5060694111 @default.
- W4287162887 hasAuthorship W4287162887A5077406767 @default.
- W4287162887 hasBestOaLocation W42871628871 @default.
- W4287162887 hasConcept C105795698 @default.
- W4287162887 hasConcept C108650721 @default.
- W4287162887 hasConcept C11671645 @default.
- W4287162887 hasConcept C119857082 @default.
- W4287162887 hasConcept C149782125 @default.
- W4287162887 hasConcept C154945302 @default.
- W4287162887 hasConcept C15744967 @default.
- W4287162887 hasConcept C190470478 @default.
- W4287162887 hasConcept C33923547 @default.
- W4287162887 hasConcept C37914503 @default.
- W4287162887 hasConcept C41008148 @default.
- W4287162887 hasConcept C71889745 @default.
- W4287162887 hasConcept C77805123 @default.
- W4287162887 hasConcept C97256817 @default.
- W4287162887 hasConceptScore W4287162887C105795698 @default.
- W4287162887 hasConceptScore W4287162887C108650721 @default.
- W4287162887 hasConceptScore W4287162887C11671645 @default.
- W4287162887 hasConceptScore W4287162887C119857082 @default.
- W4287162887 hasConceptScore W4287162887C149782125 @default.
- W4287162887 hasConceptScore W4287162887C154945302 @default.
- W4287162887 hasConceptScore W4287162887C15744967 @default.
- W4287162887 hasConceptScore W4287162887C190470478 @default.
- W4287162887 hasConceptScore W4287162887C33923547 @default.
- W4287162887 hasConceptScore W4287162887C37914503 @default.
- W4287162887 hasConceptScore W4287162887C41008148 @default.
- W4287162887 hasConceptScore W4287162887C71889745 @default.
- W4287162887 hasConceptScore W4287162887C77805123 @default.
- W4287162887 hasConceptScore W4287162887C97256817 @default.
- W4287162887 hasLocation W42871628871 @default.
- W4287162887 hasOpenAccess W4287162887 @default.
- W4287162887 hasPrimaryLocation W42871628871 @default.
- W4287162887 hasRelatedWork W2582133354 @default.
- W4287162887 hasRelatedWork W3149487685 @default.
- W4287162887 hasRelatedWork W3171731510 @default.
- W4287162887 hasRelatedWork W3186430488 @default.
- W4287162887 hasRelatedWork W4238563872 @default.
- W4287162887 hasRelatedWork W4280530714 @default.
- W4287162887 hasRelatedWork W4283795504 @default.
- W4287162887 hasRelatedWork W4287162887 @default.
- W4287162887 hasRelatedWork W4296931829 @default.
- W4287162887 hasRelatedWork W52128503 @default.
- W4287162887 isParatext "false" @default.
- W4287162887 isRetracted "false" @default.
- W4287162887 workType "article" @default.