Matches in SemOpenAlex for { <https://semopenalex.org/work/W3036255908> ?p ?o ?g. }
- W3036255908 abstract "In unsupervised domain adaptation, existing theory focuses on situations where the source and target domains are close. In practice, conditional entropy minimization and pseudo-labeling work even when the domain shifts are much larger than those analyzed by existing theory. We identify and analyze one particular setting where the domain shift can be large, but these algorithms provably work: certain spurious features correlate with the label in the source domain but are independent of the label in the target. Our analysis considers linear classification where the spurious features are Gaussian and the non-spurious features are a mixture of log-concave distributions. For this setting, we prove that entropy minimization on unlabeled target data will avoid using the spurious feature if initialized with a decently accurate source classifier, even though the objective is non-convex and contains multiple bad local minima using the spurious features. We verify our theory for spurious domain shift tasks on semi-synthetic Celeb-A and MNIST datasets. Our results suggest that practitioners collect and self-train on large, diverse datasets to reduce biases in classifiers even if labeling is impractical." @default.
- W3036255908 created "2020-06-25" @default.
- W3036255908 creator A5014379995 @default.
- W3036255908 creator A5055799536 @default.
- W3036255908 creator A5061905935 @default.
- W3036255908 creator A5066483084 @default.
- W3036255908 date "2020-06-17" @default.
- W3036255908 modified "2023-09-27" @default.
- W3036255908 title "Self-training Avoids Using Spurious Features Under Domain Shift" @default.
- W3036255908 cites W140084236 @default.
- W3036255908 cites W1565327149 @default.
- W3036255908 cites W1589919686 @default.
- W3036255908 cites W1644055776 @default.
- W3036255908 cites W1660353176 @default.
- W3036255908 cites W1834627138 @default.
- W3036255908 cites W189742998 @default.
- W3036255908 cites W2009941369 @default.
- W3036255908 cites W2034368206 @default.
- W3036255908 cites W2048679005 @default.
- W3036255908 cites W2069735839 @default.
- W3036255908 cites W2096943734 @default.
- W3036255908 cites W2104094955 @default.
- W3036255908 cites W2112483442 @default.
- W3036255908 cites W2114718442 @default.
- W3036255908 cites W2184987264 @default.
- W3036255908 cites W2593768305 @default.
- W3036255908 cites W2609037894 @default.
- W3036255908 cites W2765345568 @default.
- W3036255908 cites W2897355816 @default.
- W3036255908 cites W2907520152 @default.
- W3036255908 cites W2951286828 @default.
- W3036255908 cites W2962970380 @default.
- W3036255908 cites W2962972504 @default.
- W3036255908 cites W2963094258 @default.
- W3036255908 cites W2963350032 @default.
- W3036255908 cites W2963826681 @default.
- W3036255908 cites W2964084001 @default.
- W3036255908 cites W2964106499 @default.
- W3036255908 cites W2964159205 @default.
- W3036255908 cites W2970115835 @default.
- W3036255908 cites W2970680991 @default.
- W3036255908 cites W2971316968 @default.
- W3036255908 cites W2985406498 @default.
- W3036255908 cites W2996501936 @default.
- W3036255908 cites W3001197829 @default.
- W3036255908 cites W3007380374 @default.
- W3036255908 cites W3021336872 @default.
- W3036255908 cites W3035160371 @default.
- W3036255908 cites W3108961219 @default.
- W3036255908 hasPublicationYear "2020" @default.
- W3036255908 type Work @default.
- W3036255908 sameAs 3036255908 @default.
- W3036255908 citedByCount "1" @default.
- W3036255908 countsByYear W30362559082021 @default.
- W3036255908 crossrefType "posted-content" @default.
- W3036255908 hasAuthorship W3036255908A5014379995 @default.
- W3036255908 hasAuthorship W3036255908A5055799536 @default.
- W3036255908 hasAuthorship W3036255908A5061905935 @default.
- W3036255908 hasAuthorship W3036255908A5066483084 @default.
- W3036255908 hasConcept C106301342 @default.
- W3036255908 hasConcept C11413529 @default.
- W3036255908 hasConcept C119857082 @default.
- W3036255908 hasConcept C121332964 @default.
- W3036255908 hasConcept C134306372 @default.
- W3036255908 hasConcept C147764199 @default.
- W3036255908 hasConcept C153180895 @default.
- W3036255908 hasConcept C154945302 @default.
- W3036255908 hasConcept C163716315 @default.
- W3036255908 hasConcept C186633575 @default.
- W3036255908 hasConcept C190502265 @default.
- W3036255908 hasConcept C199360897 @default.
- W3036255908 hasConcept C2776434776 @default.
- W3036255908 hasConcept C33923547 @default.
- W3036255908 hasConcept C36503486 @default.
- W3036255908 hasConcept C41008148 @default.
- W3036255908 hasConcept C50644808 @default.
- W3036255908 hasConcept C62520636 @default.
- W3036255908 hasConcept C95623464 @default.
- W3036255908 hasConcept C97256817 @default.
- W3036255908 hasConceptScore W3036255908C106301342 @default.
- W3036255908 hasConceptScore W3036255908C11413529 @default.
- W3036255908 hasConceptScore W3036255908C119857082 @default.
- W3036255908 hasConceptScore W3036255908C121332964 @default.
- W3036255908 hasConceptScore W3036255908C134306372 @default.
- W3036255908 hasConceptScore W3036255908C147764199 @default.
- W3036255908 hasConceptScore W3036255908C153180895 @default.
- W3036255908 hasConceptScore W3036255908C154945302 @default.
- W3036255908 hasConceptScore W3036255908C163716315 @default.
- W3036255908 hasConceptScore W3036255908C186633575 @default.
- W3036255908 hasConceptScore W3036255908C190502265 @default.
- W3036255908 hasConceptScore W3036255908C199360897 @default.
- W3036255908 hasConceptScore W3036255908C2776434776 @default.
- W3036255908 hasConceptScore W3036255908C33923547 @default.
- W3036255908 hasConceptScore W3036255908C36503486 @default.
- W3036255908 hasConceptScore W3036255908C41008148 @default.
- W3036255908 hasConceptScore W3036255908C50644808 @default.
- W3036255908 hasConceptScore W3036255908C62520636 @default.
- W3036255908 hasConceptScore W3036255908C95623464 @default.
- W3036255908 hasConceptScore W3036255908C97256817 @default.
- W3036255908 hasLocation W30362559081 @default.