Matches in SemOpenAlex for { <https://semopenalex.org/work/W2284050935> ?p ?o ?g. }
- W2284050935 abstract "We present weight normalization: a reparameterization of the weight vectors in a neural network that decouples the length of those weight vectors from their direction. By reparameterizing the weights in this way we improve the conditioning of the optimization problem and we speed up convergence of stochastic gradient descent. Our reparameterization is inspired by batch normalization but does not introduce any dependencies between the examples in a minibatch. This means that our method can also be applied successfully to recurrent models such as LSTMs and to noise-sensitive applications such as deep reinforcement learning or generative models, for which batch normalization is less well suited. Although our method is much simpler, it still provides much of the speed-up of full batch normalization. In addition, the computational overhead of our method is lower, permitting more optimization steps to be taken in the same amount of time. We demonstrate the usefulness of our method on applications in supervised image recognition, generative modelling, and deep reinforcement learning." @default.
- W2284050935 created "2016-06-24" @default.
- W2284050935 creator A5024065815 @default.
- W2284050935 creator A5082512329 @default.
- W2284050935 date "2016-02-25" @default.
- W2284050935 modified "2023-09-30" @default.
- W2284050935 title "Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks" @default.
- W2284050935 cites W104184427 @default.
- W2284050935 cites W1492459858 @default.
- W2284050935 cites W1522301498 @default.
- W2284050935 cites W1606347560 @default.
- W2284050935 cites W1850742715 @default.
- W2284050935 cites W1877062207 @default.
- W2284050935 cites W1906598733 @default.
- W2284050935 cites W1909320841 @default.
- W2284050935 cites W1915968771 @default.
- W2284050935 cites W196761320 @default.
- W2284050935 cites W2064675550 @default.
- W2284050935 cites W2086161653 @default.
- W2284050935 cites W2145339207 @default.
- W2284050935 cites W2152424459 @default.
- W2284050935 cites W2155894447 @default.
- W2284050935 cites W2167729035 @default.
- W2284050935 cites W2168894214 @default.
- W2284050935 cites W2178237821 @default.
- W2284050935 cites W2194775991 @default.
- W2284050935 cites W2287334441 @default.
- W2284050935 cites W2949117887 @default.
- W2284050935 cites W2951004968 @default.
- W2284050935 cites W2963382180 @default.
- W2284050935 cites W2963911037 @default.
- W2284050935 cites W3037950864 @default.
- W2284050935 cites W3118608800 @default.
- W2284050935 hasPublicationYear "2016" @default.
- W2284050935 type Work @default.
- W2284050935 sameAs 2284050935 @default.
- W2284050935 citedByCount "172" @default.
- W2284050935 countsByYear W22840509352016 @default.
- W2284050935 countsByYear W22840509352017 @default.
- W2284050935 countsByYear W22840509352018 @default.
- W2284050935 countsByYear W22840509352019 @default.
- W2284050935 countsByYear W22840509352020 @default.
- W2284050935 countsByYear W22840509352021 @default.
- W2284050935 countsByYear W22840509352022 @default.
- W2284050935 crossrefType "posted-content" @default.
- W2284050935 hasAuthorship W2284050935A5024065815 @default.
- W2284050935 hasAuthorship W2284050935A5082512329 @default.
- W2284050935 hasConcept C108583219 @default.
- W2284050935 hasConcept C11413529 @default.
- W2284050935 hasConcept C119857082 @default.
- W2284050935 hasConcept C136886441 @default.
- W2284050935 hasConcept C144024400 @default.
- W2284050935 hasConcept C153258448 @default.
- W2284050935 hasConcept C154945302 @default.
- W2284050935 hasConcept C167966045 @default.
- W2284050935 hasConcept C19165224 @default.
- W2284050935 hasConcept C206688291 @default.
- W2284050935 hasConcept C2984842247 @default.
- W2284050935 hasConcept C39890363 @default.
- W2284050935 hasConcept C41008148 @default.
- W2284050935 hasConcept C50644808 @default.
- W2284050935 hasConcept C97541855 @default.
- W2284050935 hasConceptScore W2284050935C108583219 @default.
- W2284050935 hasConceptScore W2284050935C11413529 @default.
- W2284050935 hasConceptScore W2284050935C119857082 @default.
- W2284050935 hasConceptScore W2284050935C136886441 @default.
- W2284050935 hasConceptScore W2284050935C144024400 @default.
- W2284050935 hasConceptScore W2284050935C153258448 @default.
- W2284050935 hasConceptScore W2284050935C154945302 @default.
- W2284050935 hasConceptScore W2284050935C167966045 @default.
- W2284050935 hasConceptScore W2284050935C19165224 @default.
- W2284050935 hasConceptScore W2284050935C206688291 @default.
- W2284050935 hasConceptScore W2284050935C2984842247 @default.
- W2284050935 hasConceptScore W2284050935C39890363 @default.
- W2284050935 hasConceptScore W2284050935C41008148 @default.
- W2284050935 hasConceptScore W2284050935C50644808 @default.
- W2284050935 hasConceptScore W2284050935C97541855 @default.
- W2284050935 hasLocation W22840509351 @default.
- W2284050935 hasOpenAccess W2284050935 @default.
- W2284050935 hasPrimaryLocation W22840509351 @default.
- W2284050935 hasRelatedWork W1533861849 @default.
- W2284050935 hasRelatedWork W1677182931 @default.
- W2284050935 hasRelatedWork W1686810756 @default.
- W2284050935 hasRelatedWork W2064675550 @default.
- W2284050935 hasRelatedWork W2095705004 @default.
- W2284050935 hasRelatedWork W2097117768 @default.
- W2284050935 hasRelatedWork W2099471712 @default.
- W2284050935 hasRelatedWork W2108598243 @default.
- W2284050935 hasRelatedWork W2112796928 @default.
- W2284050935 hasRelatedWork W2117539524 @default.
- W2284050935 hasRelatedWork W2146502635 @default.
- W2284050935 hasRelatedWork W2163605009 @default.
- W2284050935 hasRelatedWork W2173520492 @default.
- W2284050935 hasRelatedWork W2194775991 @default.
- W2284050935 hasRelatedWork W2432004435 @default.
- W2284050935 hasRelatedWork W2502312327 @default.
- W2284050935 hasRelatedWork W2949117887 @default.
- W2284050935 hasRelatedWork W2963446712 @default.
- W2284050935 hasRelatedWork W2964121744 @default.
- W2284050935 hasRelatedWork W3118608800 @default.