Matches in SemOpenAlex for { <https://semopenalex.org/work/W3216093740> ?p ?o ?g. }
- W3216093740 abstract "We argue that a valuable perspective on when a model learns textit{good} representations is that inputs that are mapped to similar representations by the model should be perceived similarly by humans. We use textit{representation inversion} to generate multiple inputs that map to the same model representation, then quantify the perceptual similarity of these inputs via human surveys. Our approach yields a measure of the extent to which a model is aligned with human perception. Using this measure of alignment, we evaluate models trained with various learning paradigms (eg~supervised and self-supervised learning) and different training losses (standard and robust training). Our results suggest that the alignment of representations with human perception provides useful additional insights into the qualities of a model. For example, we find that alignment with human perception can be used as a measure of trust in a model's prediction on inputs where different models have conflicting outputs. We also find that various properties of a model like its architecture, training paradigm, training loss, and data augmentation play a significant role in learning representations that are aligned with human perception." @default.
- W3216093740 created "2021-12-06" @default.
- W3216093740 creator A5037961350 @default.
- W3216093740 creator A5042278493 @default.
- W3216093740 creator A5048896019 @default.
- W3216093740 creator A5051069133 @default.
- W3216093740 creator A5067688305 @default.
- W3216093740 creator A5069362336 @default.
- W3216093740 creator A5071538560 @default.
- W3216093740 date "2021-11-29" @default.
- W3216093740 modified "2023-09-23" @default.
- W3216093740 title "Exploring Alignment of Representations with Human Perception" @default.
- W3216093740 cites W1825675169 @default.
- W3216093740 cites W1849277567 @default.
- W3216093740 cites W1883420340 @default.
- W3216093740 cites W1945616565 @default.
- W3216093740 cites W2002995732 @default.
- W3216093740 cites W2009284521 @default.
- W3216093740 cites W2178314882 @default.
- W3216093740 cites W2273348943 @default.
- W3216093740 cites W2576670953 @default.
- W3216093740 cites W2586275201 @default.
- W3216093740 cites W2783879794 @default.
- W3216093740 cites W2787708942 @default.
- W3216093740 cites W2795373741 @default.
- W3216093740 cites W2796885425 @default.
- W3216093740 cites W2807716166 @default.
- W3216093740 cites W2885106262 @default.
- W3216093740 cites W2898482221 @default.
- W3216093740 cites W2913266441 @default.
- W3216093740 cites W2913848079 @default.
- W3216093740 cites W2919115771 @default.
- W3216093740 cites W2949293387 @default.
- W3216093740 cites W2949558627 @default.
- W3216093740 cites W2949987032 @default.
- W3216093740 cites W2951792921 @default.
- W3216093740 cites W2963744840 @default.
- W3216093740 cites W2964153729 @default.
- W3216093740 cites W2970317235 @default.
- W3216093740 cites W2980620660 @default.
- W3216093740 cites W2982584252 @default.
- W3216093740 cites W2991497298 @default.
- W3216093740 cites W2999405036 @default.
- W3216093740 cites W3005680577 @default.
- W3216093740 cites W3007305010 @default.
- W3216093740 cites W3010694149 @default.
- W3216093740 cites W3034318565 @default.
- W3216093740 cites W3093096163 @default.
- W3216093740 cites W3104911444 @default.
- W3216093740 cites W3108490213 @default.
- W3216093740 cites W3175337229 @default.
- W3216093740 hasPublicationYear "2021" @default.
- W3216093740 type Work @default.
- W3216093740 sameAs 3216093740 @default.
- W3216093740 citedByCount "0" @default.
- W3216093740 crossrefType "posted-content" @default.
- W3216093740 hasAuthorship W3216093740A5037961350 @default.
- W3216093740 hasAuthorship W3216093740A5042278493 @default.
- W3216093740 hasAuthorship W3216093740A5048896019 @default.
- W3216093740 hasAuthorship W3216093740A5051069133 @default.
- W3216093740 hasAuthorship W3216093740A5067688305 @default.
- W3216093740 hasAuthorship W3216093740A5069362336 @default.
- W3216093740 hasAuthorship W3216093740A5071538560 @default.
- W3216093740 hasConcept C103278499 @default.
- W3216093740 hasConcept C109007969 @default.
- W3216093740 hasConcept C115961682 @default.
- W3216093740 hasConcept C119857082 @default.
- W3216093740 hasConcept C124101348 @default.
- W3216093740 hasConcept C12713177 @default.
- W3216093740 hasConcept C151730666 @default.
- W3216093740 hasConcept C154945302 @default.
- W3216093740 hasConcept C15744967 @default.
- W3216093740 hasConcept C169760540 @default.
- W3216093740 hasConcept C17744445 @default.
- W3216093740 hasConcept C1893757 @default.
- W3216093740 hasConcept C199539241 @default.
- W3216093740 hasConcept C26760741 @default.
- W3216093740 hasConcept C2776359362 @default.
- W3216093740 hasConcept C2780009758 @default.
- W3216093740 hasConcept C41008148 @default.
- W3216093740 hasConcept C86803240 @default.
- W3216093740 hasConcept C94625758 @default.
- W3216093740 hasConceptScore W3216093740C103278499 @default.
- W3216093740 hasConceptScore W3216093740C109007969 @default.
- W3216093740 hasConceptScore W3216093740C115961682 @default.
- W3216093740 hasConceptScore W3216093740C119857082 @default.
- W3216093740 hasConceptScore W3216093740C124101348 @default.
- W3216093740 hasConceptScore W3216093740C12713177 @default.
- W3216093740 hasConceptScore W3216093740C151730666 @default.
- W3216093740 hasConceptScore W3216093740C154945302 @default.
- W3216093740 hasConceptScore W3216093740C15744967 @default.
- W3216093740 hasConceptScore W3216093740C169760540 @default.
- W3216093740 hasConceptScore W3216093740C17744445 @default.
- W3216093740 hasConceptScore W3216093740C1893757 @default.
- W3216093740 hasConceptScore W3216093740C199539241 @default.
- W3216093740 hasConceptScore W3216093740C26760741 @default.
- W3216093740 hasConceptScore W3216093740C2776359362 @default.
- W3216093740 hasConceptScore W3216093740C2780009758 @default.
- W3216093740 hasConceptScore W3216093740C41008148 @default.
- W3216093740 hasConceptScore W3216093740C86803240 @default.