Matches in SemOpenAlex for { <https://semopenalex.org/work/W3099360433> ?p ?o ?g. }
Showing items 1 to 97 of
97
with 100 items per page.
- W3099360433 abstract "Modern deep learning is primarily an experimental science, in which empirical advances occasionally come at the expense of probabilistic rigor. Here we focus on one such example; namely the use of the categorical cross-entropy loss to model data that is not strictly categorical, but rather takes values on the simplex. This practice is standard in neural network architectures with label smoothing and actor-mimic reinforcement learning, amongst others. Drawing on the recently discovered continuous-categorical distribution, we propose probabilistically-inspired alternatives to these models, providing an approach that is more principled and theoretically appealing. Through careful experimentation, including an ablation study, we identify the potential for outperformance in these models, thereby highlighting the importance of a proper probabilistic treatment, as well as illustrating some of the failure modes thereof." @default.
- W3099360433 created "2020-11-23" @default.
- W3099360433 creator A5020477291 @default.
- W3099360433 creator A5030697996 @default.
- W3099360433 creator A5065203625 @default.
- W3099360433 creator A5065923620 @default.
- W3099360433 date "2020-11-10" @default.
- W3099360433 modified "2023-09-27" @default.
- W3099360433 title "Uses and Abuses of the Cross-Entropy Loss: Case Studies in Modern Deep Learning." @default.
- W3099360433 cites W1513838499 @default.
- W3099360433 cites W1677182931 @default.
- W3099360433 cites W1821462560 @default.
- W3099360433 cites W2078112764 @default.
- W3099360433 cites W2111406701 @default.
- W3099360433 cites W2144416900 @default.
- W3099360433 cites W2144513243 @default.
- W3099360433 cites W2145339207 @default.
- W3099360433 cites W2174786457 @default.
- W3099360433 cites W2183341477 @default.
- W3099360433 cites W2186488806 @default.
- W3099360433 cites W2214409633 @default.
- W3099360433 cites W2626017178 @default.
- W3099360433 cites W2787919999 @default.
- W3099360433 cites W2904250115 @default.
- W3099360433 cites W2942801205 @default.
- W3099360433 cites W2945313542 @default.
- W3099360433 cites W2949117887 @default.
- W3099360433 cites W2950903920 @default.
- W3099360433 cites W2954996726 @default.
- W3099360433 cites W2962734576 @default.
- W3099360433 cites W2963403868 @default.
- W3099360433 cites W2964081807 @default.
- W3099360433 cites W2965658867 @default.
- W3099360433 cites W2970206392 @default.
- W3099360433 cites W2971246770 @default.
- W3099360433 cites W3034825847 @default.
- W3099360433 hasPublicationYear "2020" @default.
- W3099360433 type Work @default.
- W3099360433 sameAs 3099360433 @default.
- W3099360433 citedByCount "2" @default.
- W3099360433 countsByYear W30993604332021 @default.
- W3099360433 crossrefType "posted-content" @default.
- W3099360433 hasAuthorship W3099360433A5020477291 @default.
- W3099360433 hasAuthorship W3099360433A5030697996 @default.
- W3099360433 hasAuthorship W3099360433A5065203625 @default.
- W3099360433 hasAuthorship W3099360433A5065923620 @default.
- W3099360433 hasConcept C106301342 @default.
- W3099360433 hasConcept C108583219 @default.
- W3099360433 hasConcept C119857082 @default.
- W3099360433 hasConcept C121332964 @default.
- W3099360433 hasConcept C154945302 @default.
- W3099360433 hasConcept C167981619 @default.
- W3099360433 hasConcept C41008148 @default.
- W3099360433 hasConcept C49937458 @default.
- W3099360433 hasConcept C5274069 @default.
- W3099360433 hasConcept C62520636 @default.
- W3099360433 hasConcept C9679016 @default.
- W3099360433 hasConcept C97541855 @default.
- W3099360433 hasConceptScore W3099360433C106301342 @default.
- W3099360433 hasConceptScore W3099360433C108583219 @default.
- W3099360433 hasConceptScore W3099360433C119857082 @default.
- W3099360433 hasConceptScore W3099360433C121332964 @default.
- W3099360433 hasConceptScore W3099360433C154945302 @default.
- W3099360433 hasConceptScore W3099360433C167981619 @default.
- W3099360433 hasConceptScore W3099360433C41008148 @default.
- W3099360433 hasConceptScore W3099360433C49937458 @default.
- W3099360433 hasConceptScore W3099360433C5274069 @default.
- W3099360433 hasConceptScore W3099360433C62520636 @default.
- W3099360433 hasConceptScore W3099360433C9679016 @default.
- W3099360433 hasConceptScore W3099360433C97541855 @default.
- W3099360433 hasLocation W30993604331 @default.
- W3099360433 hasOpenAccess W3099360433 @default.
- W3099360433 hasPrimaryLocation W30993604331 @default.
- W3099360433 hasRelatedWork W2142964748 @default.
- W3099360433 hasRelatedWork W2160805123 @default.
- W3099360433 hasRelatedWork W2463882827 @default.
- W3099360433 hasRelatedWork W2471268222 @default.
- W3099360433 hasRelatedWork W2522500382 @default.
- W3099360433 hasRelatedWork W2889175957 @default.
- W3099360433 hasRelatedWork W2941291441 @default.
- W3099360433 hasRelatedWork W2946940672 @default.
- W3099360433 hasRelatedWork W2961430136 @default.
- W3099360433 hasRelatedWork W2963415369 @default.
- W3099360433 hasRelatedWork W2970534725 @default.
- W3099360433 hasRelatedWork W3038853105 @default.
- W3099360433 hasRelatedWork W3091373078 @default.
- W3099360433 hasRelatedWork W3121305982 @default.
- W3099360433 hasRelatedWork W3167787620 @default.
- W3099360433 hasRelatedWork W3182734776 @default.
- W3099360433 hasRelatedWork W3194140983 @default.
- W3099360433 hasRelatedWork W3209807813 @default.
- W3099360433 hasRelatedWork W79523873 @default.
- W3099360433 hasRelatedWork W807002533 @default.
- W3099360433 isParatext "false" @default.
- W3099360433 isRetracted "false" @default.
- W3099360433 magId "3099360433" @default.
- W3099360433 workType "article" @default.