Matches in SemOpenAlex for { <https://semopenalex.org/work/W3041629653> ?p ?o ?g. }
- W3041629653 abstract "Many common loss functions such as mean-squared-error, cross-entropy, and reconstruction loss are unnecessarily rigid. Under a probabilistic interpretation, these common losses correspond to distributions with fixed shapes and scales. We instead argue for optimizing full likelihoods that include parameters like the normal variance and softmax temperature. Joint optimization of these ``likelihood parameters'' with model parameters can adaptively tune the scales and shapes of losses in addition to the strength of regularization. We explore and systematically evaluate how to parameterize and apply likelihood parameters for robust modeling, outlier-detection, and re-calibration. Additionally, we propose adaptively tuning L2 and L1 weights by fitting the scale parameters of normal and Laplace priors and introduce more flexible element-wise regularizers." @default.
- W3041629653 created "2020-07-16" @default.
- W3041629653 creator A5020277184 @default.
- W3041629653 creator A5023786468 @default.
- W3041629653 creator A5074429265 @default.
- W3041629653 date "2021-05-04" @default.
- W3041629653 modified "2023-09-27" @default.
- W3041629653 title "It Is Likely That Your Loss Should be a Likelihood" @default.
- W3041629653 cites W102279446 @default.
- W3041629653 cites W1530232915 @default.
- W3041629653 cites W1533861849 @default.
- W3041629653 cites W1554944419 @default.
- W3041629653 cites W1560021816 @default.
- W3041629653 cites W1579853615 @default.
- W3041629653 cites W1598033630 @default.
- W3041629653 cites W1618905105 @default.
- W3041629653 cites W1663973292 @default.
- W3041629653 cites W1677182931 @default.
- W3041629653 cites W1821462560 @default.
- W3041629653 cites W1834627138 @default.
- W3041629653 cites W187043655 @default.
- W3041629653 cites W1959608418 @default.
- W3041629653 cites W1980262437 @default.
- W3041629653 cites W1995443851 @default.
- W3041629653 cites W2019014808 @default.
- W3041629653 cites W2056081083 @default.
- W3041629653 cites W2062227835 @default.
- W3041629653 cites W2063978378 @default.
- W3041629653 cites W2069051541 @default.
- W3041629653 cites W2085261163 @default.
- W3041629653 cites W2101234009 @default.
- W3041629653 cites W2105497548 @default.
- W3041629653 cites W2108598243 @default.
- W3041629653 cites W2117670920 @default.
- W3041629653 cites W2122538988 @default.
- W3041629653 cites W2129281431 @default.
- W3041629653 cites W2144182447 @default.
- W3041629653 cites W2194775991 @default.
- W3041629653 cites W2254249950 @default.
- W3041629653 cites W2296719434 @default.
- W3041629653 cites W2326925005 @default.
- W3041629653 cites W2335728318 @default.
- W3041629653 cites W2336788611 @default.
- W3041629653 cites W2338990760 @default.
- W3041629653 cites W2600383743 @default.
- W3041629653 cites W2611736676 @default.
- W3041629653 cites W2801490189 @default.
- W3041629653 cites W2912934387 @default.
- W3041629653 cites W2949517790 @default.
- W3041629653 cites W2955863859 @default.
- W3041629653 cites W2963081736 @default.
- W3041629653 cites W2963445059 @default.
- W3041629653 cites W2963446712 @default.
- W3041629653 cites W2963677766 @default.
- W3041629653 cites W2964121744 @default.
- W3041629653 cites W2964212410 @default.
- W3041629653 cites W2991349609 @default.
- W3041629653 cites W3118608800 @default.
- W3041629653 cites W3120740533 @default.
- W3041629653 cites W3128465814 @default.
- W3041629653 hasPublicationYear "2021" @default.
- W3041629653 type Work @default.
- W3041629653 sameAs 3041629653 @default.
- W3041629653 citedByCount "0" @default.
- W3041629653 crossrefType "journal-article" @default.
- W3041629653 hasAuthorship W3041629653A5020277184 @default.
- W3041629653 hasAuthorship W3041629653A5023786468 @default.
- W3041629653 hasAuthorship W3041629653A5074429265 @default.
- W3041629653 hasConcept C105795698 @default.
- W3041629653 hasConcept C107673813 @default.
- W3041629653 hasConcept C111472728 @default.
- W3041629653 hasConcept C11413529 @default.
- W3041629653 hasConcept C121955636 @default.
- W3041629653 hasConcept C138885662 @default.
- W3041629653 hasConcept C144133560 @default.
- W3041629653 hasConcept C154945302 @default.
- W3041629653 hasConcept C177769412 @default.
- W3041629653 hasConcept C188441871 @default.
- W3041629653 hasConcept C196083921 @default.
- W3041629653 hasConcept C2776135515 @default.
- W3041629653 hasConcept C28826006 @default.
- W3041629653 hasConcept C33923547 @default.
- W3041629653 hasConcept C41008148 @default.
- W3041629653 hasConcept C50644808 @default.
- W3041629653 hasConcept C75553542 @default.
- W3041629653 hasConcept C79337645 @default.
- W3041629653 hasConceptScore W3041629653C105795698 @default.
- W3041629653 hasConceptScore W3041629653C107673813 @default.
- W3041629653 hasConceptScore W3041629653C111472728 @default.
- W3041629653 hasConceptScore W3041629653C11413529 @default.
- W3041629653 hasConceptScore W3041629653C121955636 @default.
- W3041629653 hasConceptScore W3041629653C138885662 @default.
- W3041629653 hasConceptScore W3041629653C144133560 @default.
- W3041629653 hasConceptScore W3041629653C154945302 @default.
- W3041629653 hasConceptScore W3041629653C177769412 @default.
- W3041629653 hasConceptScore W3041629653C188441871 @default.
- W3041629653 hasConceptScore W3041629653C196083921 @default.
- W3041629653 hasConceptScore W3041629653C2776135515 @default.
- W3041629653 hasConceptScore W3041629653C28826006 @default.
- W3041629653 hasConceptScore W3041629653C33923547 @default.