Matches in SemOpenAlex for { <https://semopenalex.org/work/W2808172644> ?p ?o ?g. }
- W2808172644 abstract "We focus on two supervised visual reasoning tasks whose labels encode a semantic relational rule between two or more objects in an image: the MNIST Parity task and the colorized Pentomino task. The objects in the images undergo random translation, scaling, rotation and coloring transformations. Thus these tasks involve invariant relational reasoning. We report uneven performance of various deep CNN models on these two tasks. For the MNIST Parity task, we report that the VGG19 model soundly outperforms a family of ResNet models. Moreover, the family of ResNet models exhibits a general sensitivity to random initialization for the MNIST Parity task. For the colorized Pentomino task, now both the VGG19 and ResNet models exhibit sluggish optimization and very poor test generalization, hovering around 30% test error. The CNN we tested all learn hierarchies of fully distributed features and thus encode the distributed representation prior. We are motivated by a hypothesis from cognitive neuroscience which posits that the human visual cortex is modularized, and this allows the visual cortex to learn higher order invariances. To this end, we consider a modularized variant of the ResNet model, referred to as a Residual Mixture Network (ResMixNet) which employs a mixture-of-experts architecture to interleave distributed representations with more specialized, modular representations. We show that very shallow ResMixNets are capable of learning each of the two tasks well, attaining less than 2% and 1% test error on the MNIST Parity and the colorized Pentomino tasks respectively. Most importantly, the ResMixNet models are extremely parameter efficient: generalizing better than various non-modular CNNs that have over 10x the number of parameters. These experimental results support the hypothesis that modularity is a robust prior for learning invariant relational reasoning." @default.
- W2808172644 created "2018-06-21" @default.
- W2808172644 creator A5061978177 @default.
- W2808172644 creator A5083701330 @default.
- W2808172644 creator A5086198262 @default.
- W2808172644 date "2018-06-18" @default.
- W2808172644 modified "2023-09-27" @default.
- W2808172644 title "Modularity Matters: Learning Invariant Relational Reasoning Tasks" @default.
- W2808172644 cites W1522301498 @default.
- W2808172644 cites W1635792279 @default.
- W2808172644 cites W1673923490 @default.
- W2808172644 cites W1686810756 @default.
- W2808172644 cites W1831445803 @default.
- W2808172644 cites W1983927101 @default.
- W2808172644 cites W1988488214 @default.
- W2808172644 cites W2005815123 @default.
- W2808172644 cites W2024169701 @default.
- W2808172644 cites W2052986837 @default.
- W2808172644 cites W2058556535 @default.
- W2808172644 cites W2063971957 @default.
- W2808172644 cites W2064222586 @default.
- W2808172644 cites W2068470708 @default.
- W2808172644 cites W2096742551 @default.
- W2808172644 cites W2102605133 @default.
- W2808172644 cites W2103431968 @default.
- W2808172644 cites W2112796928 @default.
- W2808172644 cites W2123341385 @default.
- W2808172644 cites W2134118634 @default.
- W2808172644 cites W2150884987 @default.
- W2808172644 cites W2152206997 @default.
- W2808172644 cites W2194775991 @default.
- W2808172644 cites W2256357568 @default.
- W2808172644 cites W2395611524 @default.
- W2808172644 cites W2463565445 @default.
- W2808172644 cites W2561715562 @default.
- W2808172644 cites W2593871270 @default.
- W2808172644 cites W2624614404 @default.
- W2808172644 cites W2759063673 @default.
- W2808172644 cites W2912225506 @default.
- W2808172644 cites W2949117887 @default.
- W2808172644 cites W2951619830 @default.
- W2808172644 cites W2952339051 @default.
- W2808172644 cites W2953328958 @default.
- W2808172644 cites W2963118571 @default.
- W2808172644 cites W6908809 @default.
- W2808172644 hasPublicationYear "2018" @default.
- W2808172644 type Work @default.
- W2808172644 sameAs 2808172644 @default.
- W2808172644 citedByCount "3" @default.
- W2808172644 countsByYear W28081726442019 @default.
- W2808172644 countsByYear W28081726442021 @default.
- W2808172644 crossrefType "posted-content" @default.
- W2808172644 hasAuthorship W2808172644A5061978177 @default.
- W2808172644 hasAuthorship W2808172644A5083701330 @default.
- W2808172644 hasAuthorship W2808172644A5086198262 @default.
- W2808172644 hasConcept C101468663 @default.
- W2808172644 hasConcept C104317684 @default.
- W2808172644 hasConcept C108583219 @default.
- W2808172644 hasConcept C111919701 @default.
- W2808172644 hasConcept C119857082 @default.
- W2808172644 hasConcept C154945302 @default.
- W2808172644 hasConcept C185592680 @default.
- W2808172644 hasConcept C190470478 @default.
- W2808172644 hasConcept C190502265 @default.
- W2808172644 hasConcept C33923547 @default.
- W2808172644 hasConcept C37914503 @default.
- W2808172644 hasConcept C41008148 @default.
- W2808172644 hasConcept C55493867 @default.
- W2808172644 hasConcept C66746571 @default.
- W2808172644 hasConceptScore W2808172644C101468663 @default.
- W2808172644 hasConceptScore W2808172644C104317684 @default.
- W2808172644 hasConceptScore W2808172644C108583219 @default.
- W2808172644 hasConceptScore W2808172644C111919701 @default.
- W2808172644 hasConceptScore W2808172644C119857082 @default.
- W2808172644 hasConceptScore W2808172644C154945302 @default.
- W2808172644 hasConceptScore W2808172644C185592680 @default.
- W2808172644 hasConceptScore W2808172644C190470478 @default.
- W2808172644 hasConceptScore W2808172644C190502265 @default.
- W2808172644 hasConceptScore W2808172644C33923547 @default.
- W2808172644 hasConceptScore W2808172644C37914503 @default.
- W2808172644 hasConceptScore W2808172644C41008148 @default.
- W2808172644 hasConceptScore W2808172644C55493867 @default.
- W2808172644 hasConceptScore W2808172644C66746571 @default.
- W2808172644 hasLocation W28081726441 @default.
- W2808172644 hasOpenAccess W2808172644 @default.
- W2808172644 hasPrimaryLocation W28081726441 @default.
- W2808172644 hasRelatedWork W1606671545 @default.
- W2808172644 hasRelatedWork W2605417544 @default.
- W2808172644 hasRelatedWork W2743157634 @default.
- W2808172644 hasRelatedWork W2766928247 @default.
- W2808172644 hasRelatedWork W2793195060 @default.
- W2808172644 hasRelatedWork W2893615158 @default.
- W2808172644 hasRelatedWork W2936179624 @default.
- W2808172644 hasRelatedWork W2948467580 @default.
- W2808172644 hasRelatedWork W2951081618 @default.
- W2808172644 hasRelatedWork W2979490629 @default.
- W2808172644 hasRelatedWork W2982561969 @default.
- W2808172644 hasRelatedWork W2986385672 @default.
- W2808172644 hasRelatedWork W2991368557 @default.
- W2808172644 hasRelatedWork W3034427277 @default.