Matches in SemOpenAlex for { <https://semopenalex.org/work/W3010204512> ?p ?o ?g. }
- W3010204512 abstract "Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance in many computer vision tasks over the years. However, this comes at the cost of heavy computation and memory intensive network designs, suggesting potential improvements in efficiency. Convolutional layers of CNNs partly account for such an inefficiency, as they are known to learn redundant features. In this work, we exploit this redundancy, observing it as the correlation between convolutional filters of a layer, and propose an alternative approach to reproduce it efficiently. The proposed 'LinearConv' layer learns a set of orthogonal filters, and a set of coefficients that linearly combines them to introduce a controlled redundancy. We introduce a correlation-based regularization loss to achieve such flexibility over redundancy, and control the number of parameters in turn. This is designed as a plug-and-play layer to conveniently replace a conventional convolutional layer, without any additional changes required in the network architecture or the hyperparameter settings. Our experiments verify that LinearConv models achieve a performance on-par with their counterparts, with almost a 50% reduction in parameters on average, and the same computational requirement and speed at inference." @default.
- W3010204512 created "2020-03-13" @default.
- W3010204512 creator A5058175691 @default.
- W3010204512 creator A5077247727 @default.
- W3010204512 date "2019-07-26" @default.
- W3010204512 modified "2023-09-27" @default.
- W3010204512 title "Exploiting the Redundancy in Convolutional Filters for Parameter Reduction" @default.
- W3010204512 cites W1522301498 @default.
- W3010204512 cites W1665214252 @default.
- W3010204512 cites W1686810756 @default.
- W3010204512 cites W1787735662 @default.
- W3010204512 cites W1927052826 @default.
- W3010204512 cites W1935978687 @default.
- W3010204512 cites W2095705004 @default.
- W3010204512 cites W2097117768 @default.
- W3010204512 cites W2108598243 @default.
- W3010204512 cites W2112796928 @default.
- W3010204512 cites W2117539524 @default.
- W3010204512 cites W2121775913 @default.
- W3010204512 cites W2123045220 @default.
- W3010204512 cites W2161591461 @default.
- W3010204512 cites W2163605009 @default.
- W3010204512 cites W2167215970 @default.
- W3010204512 cites W2172166488 @default.
- W3010204512 cites W2183341477 @default.
- W3010204512 cites W2194775991 @default.
- W3010204512 cites W2246516976 @default.
- W3010204512 cites W2279098554 @default.
- W3010204512 cites W2300242332 @default.
- W3010204512 cites W2302255633 @default.
- W3010204512 cites W2319920447 @default.
- W3010204512 cites W2335728318 @default.
- W3010204512 cites W2340897893 @default.
- W3010204512 cites W2401231614 @default.
- W3010204512 cites W2422848620 @default.
- W3010204512 cites W2531409750 @default.
- W3010204512 cites W2549139847 @default.
- W3010204512 cites W2553303224 @default.
- W3010204512 cites W2553902701 @default.
- W3010204512 cites W2554242204 @default.
- W3010204512 cites W2561910431 @default.
- W3010204512 cites W2605135468 @default.
- W3010204512 cites W2606722458 @default.
- W3010204512 cites W2612445135 @default.
- W3010204512 cites W2750384547 @default.
- W3010204512 cites W2785366763 @default.
- W3010204512 cites W2807128367 @default.
- W3010204512 cites W2883780447 @default.
- W3010204512 cites W2890080559 @default.
- W3010204512 cites W2899771611 @default.
- W3010204512 cites W2938458886 @default.
- W3010204512 cites W2946706208 @default.
- W3010204512 cites W2946948417 @default.
- W3010204512 cites W2949117887 @default.
- W3010204512 cites W2950967261 @default.
- W3010204512 cites W2953212265 @default.
- W3010204512 cites W2962684187 @default.
- W3010204512 cites W2962685937 @default.
- W3010204512 cites W2963114950 @default.
- W3010204512 cites W2963125010 @default.
- W3010204512 cites W2963163009 @default.
- W3010204512 cites W2963363373 @default.
- W3010204512 cites W2963420686 @default.
- W3010204512 cites W2963446712 @default.
- W3010204512 cites W2963674932 @default.
- W3010204512 cites W2964152344 @default.
- W3010204512 cites W2964299589 @default.
- W3010204512 cites W2989808579 @default.
- W3010204512 cites W3034357629 @default.
- W3010204512 cites W3034528588 @default.
- W3010204512 cites W3035414587 @default.
- W3010204512 cites W3118608800 @default.
- W3010204512 hasPublicationYear "2019" @default.
- W3010204512 type Work @default.
- W3010204512 sameAs 3010204512 @default.
- W3010204512 citedByCount "0" @default.
- W3010204512 crossrefType "posted-content" @default.
- W3010204512 hasAuthorship W3010204512A5058175691 @default.
- W3010204512 hasAuthorship W3010204512A5077247727 @default.
- W3010204512 hasConcept C111335779 @default.
- W3010204512 hasConcept C111919701 @default.
- W3010204512 hasConcept C113775141 @default.
- W3010204512 hasConcept C11413529 @default.
- W3010204512 hasConcept C152124472 @default.
- W3010204512 hasConcept C154945302 @default.
- W3010204512 hasConcept C157899210 @default.
- W3010204512 hasConcept C2524010 @default.
- W3010204512 hasConcept C2776214188 @default.
- W3010204512 hasConcept C33923547 @default.
- W3010204512 hasConcept C41008148 @default.
- W3010204512 hasConcept C45374587 @default.
- W3010204512 hasConcept C57273362 @default.
- W3010204512 hasConcept C81363708 @default.
- W3010204512 hasConcept C8642999 @default.
- W3010204512 hasConceptScore W3010204512C111335779 @default.
- W3010204512 hasConceptScore W3010204512C111919701 @default.
- W3010204512 hasConceptScore W3010204512C113775141 @default.
- W3010204512 hasConceptScore W3010204512C11413529 @default.
- W3010204512 hasConceptScore W3010204512C152124472 @default.
- W3010204512 hasConceptScore W3010204512C154945302 @default.