Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035897262> ?p ?o ?g. }
- W3035897262 abstract "In this work, we analyze the role of the network architecture in shaping the inductive bias of deep classifiers. To that end, we start by focusing on a very simple problem, i.e., classifying a class of linearly separable distributions, and show that, depending on the direction of the discriminative feature of the distribution, many state-of-the-art deep convolutional neural networks (CNNs) have a surprisingly hard time solving this simple task. We then define as neural anisotropy directions (NADs) the vectors that encapsulate the directional inductive bias of an architecture. These vectors, which are specific for each architecture and hence act as a signature, encode the preference of a network to separate the input data based on some particular features. We provide an efficient method to identify NADs for several CNN architectures and thus reveal their directional inductive biases. Furthermore, we show that, for the CIFAR-10 dataset, NADs characterize features used by CNNs to discriminate between different classes." @default.
- W3035897262 created "2020-06-25" @default.
- W3035897262 creator A5000947076 @default.
- W3035897262 creator A5016056842 @default.
- W3035897262 creator A5053991440 @default.
- W3035897262 creator A5081421595 @default.
- W3035897262 date "2020-06-17" @default.
- W3035897262 modified "2023-10-01" @default.
- W3035897262 title "Neural Anisotropy Directions." @default.
- W3035897262 cites W1836465849 @default.
- W3035897262 cites W2106411961 @default.
- W3035897262 cites W2108598243 @default.
- W3035897262 cites W2112796928 @default.
- W3035897262 cites W2145339207 @default.
- W3035897262 cites W2194775991 @default.
- W3035897262 cites W2805516822 @default.
- W3035897262 cites W2897097528 @default.
- W3035897262 cites W2904025731 @default.
- W3035897262 cites W2912713668 @default.
- W3035897262 cites W2913314773 @default.
- W3035897262 cites W2950398529 @default.
- W3035897262 cites W2962804662 @default.
- W3035897262 cites W2962835968 @default.
- W3035897262 cites W2963100491 @default.
- W3035897262 cites W2963177640 @default.
- W3035897262 cites W2963310665 @default.
- W3035897262 cites W2963343288 @default.
- W3035897262 cites W2963374479 @default.
- W3035897262 cites W2963403868 @default.
- W3035897262 cites W2963446712 @default.
- W3035897262 cites W2963798163 @default.
- W3035897262 cites W2963833878 @default.
- W3035897262 cites W2964003773 @default.
- W3035897262 cites W2964084001 @default.
- W3035897262 cites W2964088238 @default.
- W3035897262 cites W2964321699 @default.
- W3035897262 cites W2970239942 @default.
- W3035897262 cites W2970452384 @default.
- W3035897262 cites W2971028215 @default.
- W3035897262 cites W2994759459 @default.
- W3035897262 cites W3034175346 @default.
- W3035897262 cites W3098903812 @default.
- W3035897262 cites W3104548192 @default.
- W3035897262 cites W3137695714 @default.
- W3035897262 cites W36434594 @default.
- W3035897262 hasPublicationYear "2020" @default.
- W3035897262 type Work @default.
- W3035897262 sameAs 3035897262 @default.
- W3035897262 citedByCount "1" @default.
- W3035897262 countsByYear W30358972622022 @default.
- W3035897262 crossrefType "posted-content" @default.
- W3035897262 hasAuthorship W3035897262A5000947076 @default.
- W3035897262 hasAuthorship W3035897262A5016056842 @default.
- W3035897262 hasAuthorship W3035897262A5053991440 @default.
- W3035897262 hasAuthorship W3035897262A5081421595 @default.
- W3035897262 hasConcept C104317684 @default.
- W3035897262 hasConcept C108583219 @default.
- W3035897262 hasConcept C111472728 @default.
- W3035897262 hasConcept C123657996 @default.
- W3035897262 hasConcept C127413603 @default.
- W3035897262 hasConcept C134306372 @default.
- W3035897262 hasConcept C138885662 @default.
- W3035897262 hasConcept C142362112 @default.
- W3035897262 hasConcept C153180895 @default.
- W3035897262 hasConcept C153349607 @default.
- W3035897262 hasConcept C154945302 @default.
- W3035897262 hasConcept C185592680 @default.
- W3035897262 hasConcept C197352929 @default.
- W3035897262 hasConcept C201995342 @default.
- W3035897262 hasConcept C2776401178 @default.
- W3035897262 hasConcept C2780451532 @default.
- W3035897262 hasConcept C2780586882 @default.
- W3035897262 hasConcept C28006648 @default.
- W3035897262 hasConcept C2984842247 @default.
- W3035897262 hasConcept C33923547 @default.
- W3035897262 hasConcept C41008148 @default.
- W3035897262 hasConcept C41895202 @default.
- W3035897262 hasConcept C50644808 @default.
- W3035897262 hasConcept C55493867 @default.
- W3035897262 hasConcept C66746571 @default.
- W3035897262 hasConcept C70710897 @default.
- W3035897262 hasConcept C81363708 @default.
- W3035897262 hasConcept C97931131 @default.
- W3035897262 hasConceptScore W3035897262C104317684 @default.
- W3035897262 hasConceptScore W3035897262C108583219 @default.
- W3035897262 hasConceptScore W3035897262C111472728 @default.
- W3035897262 hasConceptScore W3035897262C123657996 @default.
- W3035897262 hasConceptScore W3035897262C127413603 @default.
- W3035897262 hasConceptScore W3035897262C134306372 @default.
- W3035897262 hasConceptScore W3035897262C138885662 @default.
- W3035897262 hasConceptScore W3035897262C142362112 @default.
- W3035897262 hasConceptScore W3035897262C153180895 @default.
- W3035897262 hasConceptScore W3035897262C153349607 @default.
- W3035897262 hasConceptScore W3035897262C154945302 @default.
- W3035897262 hasConceptScore W3035897262C185592680 @default.
- W3035897262 hasConceptScore W3035897262C197352929 @default.
- W3035897262 hasConceptScore W3035897262C201995342 @default.
- W3035897262 hasConceptScore W3035897262C2776401178 @default.
- W3035897262 hasConceptScore W3035897262C2780451532 @default.
- W3035897262 hasConceptScore W3035897262C2780586882 @default.