FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2737701595> ?p ?o ?g. }

• W2737701595 abstract "Bilinear pooling of Convolutional Neural Network (CNN) features [22, 23], and their compact variants [10], have been shown to be effective at fine-grained recognition, scene categorization, texture recognition, and visual question-answering tasks among others. The resulting representation captures second-order statistics of convolutional features in a translationally invariant manner. In this paper we investigate various ways of normalizing these statistics to improve their representation power. In particular we find that the matrix square-root normalization offers significant improvements and outperforms alternative schemes such as the matrix logarithm normalization when combined with elementwise square-root and l2 normalization. This improves the accuracy by 2-3% on a range of fine-grained recognition datasets leading to a new state of the art. We also investigate how the accuracy of matrix function computations effect network training and evaluation. In particular we compare against a technique for estimating matrix square-root gradients via solving a Lyapunov equation that is more numerically accurate than computing gradients via a Singular Value Decomposition (SVD). We find that while SVD gradients are numerically inaccurate the overall effect on the final accuracy is negligible once boundary cases are handled carefully. We present an alternative scheme for computing gradients that is faster and yet it offers improvements over the baseline model. Finally we show that the matrix square-root computed approximately using a few Newton iterations is just as accurate for the classification task but allows an order-of-magnitude faster GPU implementation compared to SVD decomposition." @default.
• W2737701595 created "2017-07-31" @default.
• W2737701595 creator A5052551454 @default.
• W2737701595 creator A5074826046 @default.
• W2737701595 date "2017-07-21" @default.
• W2737701595 modified "2023-09-23" @default.
• W2737701595 title "Improved Bilinear Pooling with CNNs" @default.
• W2737701595 cites W1606858007 @default.
• W2737701595 cites W1797268635 @default.
• W2737701595 cites W1826072219 @default.
• W2737701595 cites W1846799578 @default.
• W2737701595 cites W1898560071 @default.
• W2737701595 cites W1909952827 @default.
• W2737701595 cites W1966397236 @default.
• W2737701595 cites W1984755089 @default.
• W2737701595 cites W1985242550 @default.
• W2737701595 cites W1994002998 @default.
• W2737701595 cites W2012592962 @default.
• W2737701595 cites W2017726114 @default.
• W2737701595 cites W2088748973 @default.
• W2737701595 cites W2124386111 @default.
• W2737701595 cites W2146897752 @default.
• W2737701595 cites W2147238549 @default.
• W2737701595 cites W2179042386 @default.
• W2737701595 cites W2204257188 @default.
• W2737701595 cites W2294485609 @default.
• W2737701595 cites W2475287302 @default.
• W2737701595 cites W2613763509 @default.
• W2737701595 cites W2731821979 @default.
• W2737701595 cites W2740620254 @default.
• W2737701595 cites W2952047972 @default.
• W2737701595 cites W2962835968 @default.
• W2737701595 cites W2963066927 @default.
• W2737701595 cites W2963239066 @default.
• W2737701595 cites W603908379 @default.
• W2737701595 cites W78159342 @default.
• W2737701595 doi "https://doi.org/10.48550/arxiv.1707.06772" @default.
• W2737701595 hasPublicationYear "2017" @default.
• W2737701595 type Work @default.
• W2737701595 sameAs 2737701595 @default.
• W2737701595 citedByCount "21" @default.
• W2737701595 countsByYear W27377015952017 @default.
• W2737701595 countsByYear W27377015952018 @default.
• W2737701595 countsByYear W27377015952019 @default.
• W2737701595 countsByYear W27377015952020 @default.
• W2737701595 countsByYear W27377015952021 @default.
• W2737701595 crossrefType "posted-content" @default.
• W2737701595 hasAuthorship W2737701595A5052551454 @default.
• W2737701595 hasAuthorship W2737701595A5074826046 @default.
• W2737701595 hasBestOaLocation W27377015951 @default.
• W2737701595 hasConcept C11413529 @default.
• W2737701595 hasConcept C11577676 @default.
• W2737701595 hasConcept C121332964 @default.
• W2737701595 hasConcept C134306372 @default.
• W2737701595 hasConcept C136886441 @default.
• W2737701595 hasConcept C144024400 @default.
• W2737701595 hasConcept C153180895 @default.
• W2737701595 hasConcept C154945302 @default.
• W2737701595 hasConcept C158693339 @default.
• W2737701595 hasConcept C19165224 @default.
• W2737701595 hasConcept C205203396 @default.
• W2737701595 hasConcept C22789450 @default.
• W2737701595 hasConcept C2524010 @default.
• W2737701595 hasConcept C31972630 @default.
• W2737701595 hasConcept C33923547 @default.
• W2737701595 hasConcept C39927690 @default.
• W2737701595 hasConcept C41008148 @default.
• W2737701595 hasConcept C42355184 @default.
• W2737701595 hasConcept C62520636 @default.
• W2737701595 hasConcept C70437156 @default.
• W2737701595 hasConcept C81363708 @default.
• W2737701595 hasConceptScore W2737701595C11413529 @default.
• W2737701595 hasConceptScore W2737701595C11577676 @default.
• W2737701595 hasConceptScore W2737701595C121332964 @default.
• W2737701595 hasConceptScore W2737701595C134306372 @default.
• W2737701595 hasConceptScore W2737701595C136886441 @default.
• W2737701595 hasConceptScore W2737701595C144024400 @default.
• W2737701595 hasConceptScore W2737701595C153180895 @default.
• W2737701595 hasConceptScore W2737701595C154945302 @default.
• W2737701595 hasConceptScore W2737701595C158693339 @default.
• W2737701595 hasConceptScore W2737701595C19165224 @default.
• W2737701595 hasConceptScore W2737701595C205203396 @default.
• W2737701595 hasConceptScore W2737701595C22789450 @default.
• W2737701595 hasConceptScore W2737701595C2524010 @default.
• W2737701595 hasConceptScore W2737701595C31972630 @default.
• W2737701595 hasConceptScore W2737701595C33923547 @default.
• W2737701595 hasConceptScore W2737701595C39927690 @default.
• W2737701595 hasConceptScore W2737701595C41008148 @default.
• W2737701595 hasConceptScore W2737701595C42355184 @default.
• W2737701595 hasConceptScore W2737701595C62520636 @default.
• W2737701595 hasConceptScore W2737701595C70437156 @default.
• W2737701595 hasConceptScore W2737701595C81363708 @default.
• W2737701595 hasLocation W27377015951 @default.
• W2737701595 hasLocation W27377015952 @default.
• W2737701595 hasOpenAccess W2737701595 @default.
• W2737701595 hasPrimaryLocation W27377015951 @default.
• W2737701595 hasRelatedWork W2268723579 @default.
• W2737701595 hasRelatedWork W2737701595 @default.
• W2737701595 hasRelatedWork W2758063741 @default.
• W2737701595 hasRelatedWork W2792080776 @default.

• next