Matches in SemOpenAlex for { <https://semopenalex.org/work/W2803218650> ?p ?o ?g. }
Showing items 1 to 93 of
93
with 100 items per page.
- W2803218650 abstract "In this paper, we formalize the idea behind capsule nets of using a capsule vector rather than a neuron activation to predict the label of samples. To this end, we propose to learn a group of capsule subspaces onto which an input feature vector is projected. Then the lengths of resultant capsules are used to score the probability of belonging to different classes. We train such a Capsule Projection Network (CapProNet) by learning an orthogonal projection matrix for each capsule subspace, and show that each capsule subspace is updated until it contains input feature vectors corresponding to the associated class. We will also show that the capsule projection can be viewed as normalizing the multiple columns of the weight matrix simultaneously to form an orthogonal basis, which makes it more effective in incorporating novel components of input features to update capsule representations. In other words, the capsule projection can be viewed as a multi-dimensional weight normalization in capsule subspaces, where the conventional weight normalization is simply a special case of the capsule projection onto 1D lines. Only a small negligible computing overhead is incurred to train the network in low-dimensional capsule subspaces or through an alternative hyper-power iteration to estimate the normalization matrix. Experiment results on image datasets show the presented model can greatly improve the performance of the state-of-the-art ResNet backbones by $10-20%$ and that of the Densenet by $5-7%$ respectively at the same level of computing and memory expenses. The CapProNet establishes the competitive state-of-the-art performance for the family of capsule nets by significantly reducing test errors on the benchmark datasets." @default.
- W2803218650 created "2018-06-01" @default.
- W2803218650 creator A5026424547 @default.
- W2803218650 creator A5080998009 @default.
- W2803218650 creator A5082230753 @default.
- W2803218650 date "2018-05-19" @default.
- W2803218650 modified "2023-09-27" @default.
- W2803218650 title "CapProNet: Deep Feature Learning via Orthogonal Projections onto Capsule Subspaces" @default.
- W2803218650 cites W2017726114 @default.
- W2803218650 cites W2048134878 @default.
- W2803218650 cites W2194775991 @default.
- W2803218650 cites W2302255633 @default.
- W2803218650 cites W2401231614 @default.
- W2803218650 cites W2595142274 @default.
- W2803218650 cites W2797275157 @default.
- W2803218650 cites W2902319695 @default.
- W2803218650 cites W2903024257 @default.
- W2803218650 cites W2949892913 @default.
- W2803218650 cites W2963446712 @default.
- W2803218650 cites W2963703618 @default.
- W2803218650 cites W2966661 @default.
- W2803218650 hasPublicationYear "2018" @default.
- W2803218650 type Work @default.
- W2803218650 sameAs 2803218650 @default.
- W2803218650 citedByCount "3" @default.
- W2803218650 countsByYear W28032186502018 @default.
- W2803218650 countsByYear W28032186502019 @default.
- W2803218650 countsByYear W28032186502021 @default.
- W2803218650 crossrefType "posted-content" @default.
- W2803218650 hasAuthorship W2803218650A5026424547 @default.
- W2803218650 hasAuthorship W2803218650A5080998009 @default.
- W2803218650 hasAuthorship W2803218650A5082230753 @default.
- W2803218650 hasConcept C11413529 @default.
- W2803218650 hasConcept C12362212 @default.
- W2803218650 hasConcept C136886441 @default.
- W2803218650 hasConcept C144024400 @default.
- W2803218650 hasConcept C153180895 @default.
- W2803218650 hasConcept C154945302 @default.
- W2803218650 hasConcept C175694140 @default.
- W2803218650 hasConcept C19165224 @default.
- W2803218650 hasConcept C2524010 @default.
- W2803218650 hasConcept C2778778583 @default.
- W2803218650 hasConcept C32834561 @default.
- W2803218650 hasConcept C33923547 @default.
- W2803218650 hasConcept C41008148 @default.
- W2803218650 hasConcept C57493831 @default.
- W2803218650 hasConcept C59822182 @default.
- W2803218650 hasConcept C83665646 @default.
- W2803218650 hasConcept C86803240 @default.
- W2803218650 hasConceptScore W2803218650C11413529 @default.
- W2803218650 hasConceptScore W2803218650C12362212 @default.
- W2803218650 hasConceptScore W2803218650C136886441 @default.
- W2803218650 hasConceptScore W2803218650C144024400 @default.
- W2803218650 hasConceptScore W2803218650C153180895 @default.
- W2803218650 hasConceptScore W2803218650C154945302 @default.
- W2803218650 hasConceptScore W2803218650C175694140 @default.
- W2803218650 hasConceptScore W2803218650C19165224 @default.
- W2803218650 hasConceptScore W2803218650C2524010 @default.
- W2803218650 hasConceptScore W2803218650C2778778583 @default.
- W2803218650 hasConceptScore W2803218650C32834561 @default.
- W2803218650 hasConceptScore W2803218650C33923547 @default.
- W2803218650 hasConceptScore W2803218650C41008148 @default.
- W2803218650 hasConceptScore W2803218650C57493831 @default.
- W2803218650 hasConceptScore W2803218650C59822182 @default.
- W2803218650 hasConceptScore W2803218650C83665646 @default.
- W2803218650 hasConceptScore W2803218650C86803240 @default.
- W2803218650 hasLocation W28032186501 @default.
- W2803218650 hasOpenAccess W2803218650 @default.
- W2803218650 hasPrimaryLocation W28032186501 @default.
- W2803218650 hasRelatedWork W2016982198 @default.
- W2803218650 hasRelatedWork W2056783870 @default.
- W2803218650 hasRelatedWork W2072691016 @default.
- W2803218650 hasRelatedWork W2072818362 @default.
- W2803218650 hasRelatedWork W2078725553 @default.
- W2803218650 hasRelatedWork W2080282385 @default.
- W2803218650 hasRelatedWork W2165699371 @default.
- W2803218650 hasRelatedWork W2357130430 @default.
- W2803218650 hasRelatedWork W2374829798 @default.
- W2803218650 hasRelatedWork W2616165637 @default.
- W2803218650 hasRelatedWork W2891328964 @default.
- W2803218650 hasRelatedWork W2892141848 @default.
- W2803218650 hasRelatedWork W2964079212 @default.
- W2803218650 hasRelatedWork W3010186256 @default.
- W2803218650 hasRelatedWork W3163820966 @default.
- W2803218650 hasRelatedWork W3196558741 @default.
- W2803218650 hasRelatedWork W3212278476 @default.
- W2803218650 hasRelatedWork W782706175 @default.
- W2803218650 hasRelatedWork W2844035817 @default.
- W2803218650 hasRelatedWork W2862471373 @default.
- W2803218650 isParatext "false" @default.
- W2803218650 isRetracted "false" @default.
- W2803218650 magId "2803218650" @default.
- W2803218650 workType "article" @default.