Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385361004> ?p ?o ?g. }
- W4385361004 endingPage "245" @default.
- W4385361004 startingPage "222" @default.
- W4385361004 abstract "The most significant models of neuronal networks for unsupervised machine learning now include the generative adversarial networks (GAN) as a separate branch. They have recently shown some degree of success. However, there is still a major obstacle. Many different loss functions have been created to train GAN discriminators, and they all share a similar structure: a total amount of real and fake losses that are solely dependent on the real losses and the generated data respectively. A challenge with a sum equal to two losses is that the training may benefit from one loss while harming the other, as we shall demonstrate causes mode collapse and instability. First, we suggest that the current loss function may prevent learning fine details in the data while trying to contract the discriminator. In this paper, we provide a new class of loss discriminant functions that promotes the training of the discriminator with a higher frequency than that of the generator for make the discriminator more demanding compared to the generator and vice versa. We may choose appropriate weights to train the discriminator in a way that takes advantage of the GAN’s stability by using gradients of the real and fake loss parties. Our methodology may be used with any discriminator model where the loss is the total of true and fake pieces. Our approach entails appropriately adjusting the hyper-parameters in order to enhance the training of the two opposing models. The effectiveness of our loss functions on tasks involving the creation of curves and images has been demonstrated through experiments, improving the base results by a statistically significant margin on several datasets." @default.
- W4385361004 created "2023-07-29" @default.
- W4385361004 creator A5062152618 @default.
- W4385361004 creator A5068216499 @default.
- W4385361004 creator A5091281481 @default.
- W4385361004 date "2023-01-01" @default.
- W4385361004 modified "2023-09-23" @default.
- W4385361004 title "Effect of Convulsion Layers and Hyper-parameters on the Behavior of Adversarial Neural Networks" @default.
- W4385361004 cites W2546302380 @default.
- W4385361004 cites W2755577605 @default.
- W4385361004 cites W2765811365 @default.
- W4385361004 cites W2894505661 @default.
- W4385361004 cites W2899901572 @default.
- W4385361004 cites W2962793481 @default.
- W4385361004 cites W3025091470 @default.
- W4385361004 cites W3080076970 @default.
- W4385361004 cites W3096831136 @default.
- W4385361004 cites W3130140376 @default.
- W4385361004 cites W3161810785 @default.
- W4385361004 cites W3215882375 @default.
- W4385361004 cites W4205203579 @default.
- W4385361004 cites W4220955967 @default.
- W4385361004 cites W4285265243 @default.
- W4385361004 cites W4286206354 @default.
- W4385361004 cites W4292444076 @default.
- W4385361004 cites W4296225832 @default.
- W4385361004 doi "https://doi.org/10.1007/978-3-031-39386-0_11" @default.
- W4385361004 hasPublicationYear "2023" @default.
- W4385361004 type Work @default.
- W4385361004 citedByCount "0" @default.
- W4385361004 crossrefType "book-chapter" @default.
- W4385361004 hasAuthorship W4385361004A5062152618 @default.
- W4385361004 hasAuthorship W4385361004A5068216499 @default.
- W4385361004 hasAuthorship W4385361004A5091281481 @default.
- W4385361004 hasConcept C112972136 @default.
- W4385361004 hasConcept C119857082 @default.
- W4385361004 hasConcept C121332964 @default.
- W4385361004 hasConcept C14036430 @default.
- W4385361004 hasConcept C153180895 @default.
- W4385361004 hasConcept C154945302 @default.
- W4385361004 hasConcept C163258240 @default.
- W4385361004 hasConcept C166957645 @default.
- W4385361004 hasConcept C205649164 @default.
- W4385361004 hasConcept C2775924081 @default.
- W4385361004 hasConcept C2776650193 @default.
- W4385361004 hasConcept C2777212361 @default.
- W4385361004 hasConcept C2779803651 @default.
- W4385361004 hasConcept C2780992000 @default.
- W4385361004 hasConcept C37736160 @default.
- W4385361004 hasConcept C39890363 @default.
- W4385361004 hasConcept C41008148 @default.
- W4385361004 hasConcept C47446073 @default.
- W4385361004 hasConcept C50644808 @default.
- W4385361004 hasConcept C62520636 @default.
- W4385361004 hasConcept C76155785 @default.
- W4385361004 hasConcept C78397625 @default.
- W4385361004 hasConcept C78458016 @default.
- W4385361004 hasConcept C86803240 @default.
- W4385361004 hasConcept C94915269 @default.
- W4385361004 hasConceptScore W4385361004C112972136 @default.
- W4385361004 hasConceptScore W4385361004C119857082 @default.
- W4385361004 hasConceptScore W4385361004C121332964 @default.
- W4385361004 hasConceptScore W4385361004C14036430 @default.
- W4385361004 hasConceptScore W4385361004C153180895 @default.
- W4385361004 hasConceptScore W4385361004C154945302 @default.
- W4385361004 hasConceptScore W4385361004C163258240 @default.
- W4385361004 hasConceptScore W4385361004C166957645 @default.
- W4385361004 hasConceptScore W4385361004C205649164 @default.
- W4385361004 hasConceptScore W4385361004C2775924081 @default.
- W4385361004 hasConceptScore W4385361004C2776650193 @default.
- W4385361004 hasConceptScore W4385361004C2777212361 @default.
- W4385361004 hasConceptScore W4385361004C2779803651 @default.
- W4385361004 hasConceptScore W4385361004C2780992000 @default.
- W4385361004 hasConceptScore W4385361004C37736160 @default.
- W4385361004 hasConceptScore W4385361004C39890363 @default.
- W4385361004 hasConceptScore W4385361004C41008148 @default.
- W4385361004 hasConceptScore W4385361004C47446073 @default.
- W4385361004 hasConceptScore W4385361004C50644808 @default.
- W4385361004 hasConceptScore W4385361004C62520636 @default.
- W4385361004 hasConceptScore W4385361004C76155785 @default.
- W4385361004 hasConceptScore W4385361004C78397625 @default.
- W4385361004 hasConceptScore W4385361004C78458016 @default.
- W4385361004 hasConceptScore W4385361004C86803240 @default.
- W4385361004 hasConceptScore W4385361004C94915269 @default.
- W4385361004 hasLocation W43853610041 @default.
- W4385361004 hasOpenAccess W4385361004 @default.
- W4385361004 hasPrimaryLocation W43853610041 @default.
- W4385361004 hasRelatedWork W2554314924 @default.
- W4385361004 hasRelatedWork W2561036008 @default.
- W4385361004 hasRelatedWork W2809882560 @default.
- W4385361004 hasRelatedWork W2914998939 @default.
- W4385361004 hasRelatedWork W2952936466 @default.
- W4385361004 hasRelatedWork W2953246223 @default.
- W4385361004 hasRelatedWork W3005996785 @default.
- W4385361004 hasRelatedWork W4288624664 @default.
- W4385361004 hasRelatedWork W4293320219 @default.
- W4385361004 hasRelatedWork W4328029048 @default.
- W4385361004 isParatext "false" @default.