A simple neural network module for relational reasoning. Santoro, A., Raposo, D., Barrett, D., G., Malinowski, M., Pascanu, R., Battaglia, P., & Lillicrap, T. Advances in Neural Information Processing Systems, 2017-Decem(Nips):4968-4977, 2017.
![A simple neural network module for relational reasoning [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper abstract bibtex
Paper abstract bibtex Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle relational reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.
@article{
title = {A simple neural network module for relational reasoning},
type = {article},
year = {2017},
pages = {4968-4977},
volume = {2017-Decem},
id = {6308a55b-5aaf-3182-bf5d-6af502f206f5},
created = {2021-07-12T10:19:36.485Z},
file_attached = {true},
profile_id = {ad172e55-c0e8-3aa4-8465-09fac4d5f5c8},
group_id = {1ff583c0-be37-34fa-9c04-73c69437d354},
last_modified = {2021-07-12T10:19:51.044Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
folder_uuids = {20ccb950-fef9-4ee1-800c-a60ba9f1df16},
private_publication = {false},
abstract = {Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle relational reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.},
bibtype = {article},
author = {Santoro, Adam and Raposo, David and Barrett, David G.T. and Malinowski, Mateusz and Pascanu, Razvan and Battaglia, Peter and Lillicrap, Timothy},
journal = {Advances in Neural Information Processing Systems},
number = {Nips}
}Downloads: 0
{"_id":"JFoW7aSWqryzkynsb","bibbaseid":"santoro-raposo-barrett-malinowski-pascanu-battaglia-lillicrap-asimpleneuralnetworkmoduleforrelationalreasoning-2017","authorIDs":[],"author_short":["Santoro, A.","Raposo, D.","Barrett, D., G.","Malinowski, M.","Pascanu, R.","Battaglia, P.","Lillicrap, T."],"bibdata":{"title":"A simple neural network module for relational reasoning","type":"article","year":"2017","pages":"4968-4977","volume":"2017-Decem","id":"6308a55b-5aaf-3182-bf5d-6af502f206f5","created":"2021-07-12T10:19:36.485Z","file_attached":"true","profile_id":"ad172e55-c0e8-3aa4-8465-09fac4d5f5c8","group_id":"1ff583c0-be37-34fa-9c04-73c69437d354","last_modified":"2021-07-12T10:19:51.044Z","read":false,"starred":false,"authored":false,"confirmed":"true","hidden":false,"folder_uuids":"20ccb950-fef9-4ee1-800c-a60ba9f1df16","private_publication":false,"abstract":"Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle relational reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.","bibtype":"article","author":"Santoro, Adam and Raposo, David and Barrett, David G.T. and Malinowski, Mateusz and Pascanu, Razvan and Battaglia, Peter and Lillicrap, Timothy","journal":"Advances in Neural Information Processing Systems","number":"Nips","bibtex":"@article{\n title = {A simple neural network module for relational reasoning},\n type = {article},\n year = {2017},\n pages = {4968-4977},\n volume = {2017-Decem},\n id = {6308a55b-5aaf-3182-bf5d-6af502f206f5},\n created = {2021-07-12T10:19:36.485Z},\n file_attached = {true},\n profile_id = {ad172e55-c0e8-3aa4-8465-09fac4d5f5c8},\n group_id = {1ff583c0-be37-34fa-9c04-73c69437d354},\n last_modified = {2021-07-12T10:19:51.044Z},\n read = {false},\n starred = {false},\n authored = {false},\n confirmed = {true},\n hidden = {false},\n folder_uuids = {20ccb950-fef9-4ee1-800c-a60ba9f1df16},\n private_publication = {false},\n abstract = {Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Thus, by simply augmenting convolutions, LSTMs, and MLPs with RNs, we can remove computational burden from network components that are not well-suited to handle relational reasoning, reduce overall network complexity, and gain a general ability to reason about the relations between entities and their properties.},\n bibtype = {article},\n author = {Santoro, Adam and Raposo, David and Barrett, David G.T. and Malinowski, Mateusz and Pascanu, Razvan and Battaglia, Peter and Lillicrap, Timothy},\n journal = {Advances in Neural Information Processing Systems},\n number = {Nips}\n}","author_short":["Santoro, A.","Raposo, D.","Barrett, D., G.","Malinowski, M.","Pascanu, R.","Battaglia, P.","Lillicrap, T."],"urls":{"Paper":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c/file/29c94a8f-7c81-cb69-3812-67f94d7adaf2/NIPS_2017_a_simple_neural_network_module_for_relational_reasoning_Paper.pdf.pdf"},"biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","bibbaseid":"santoro-raposo-barrett-malinowski-pascanu-battaglia-lillicrap-asimpleneuralnetworkmoduleforrelationalreasoning-2017","role":"author","metadata":{"authorlinks":{}},"downloads":0},"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","creationDate":"2020-01-27T02:13:34.106Z","downloads":0,"keywords":[],"search_terms":["simple","neural","network","module","relational","reasoning","santoro","raposo","barrett","malinowski","pascanu","battaglia","lillicrap"],"title":"A simple neural network module for relational reasoning","year":2017,"dataSources":["hEoKh4ygEAWbAZ5iy","rQFxZQs78YQJ9m34s","aXmRAq63YsH7a3ufx","ya2CyA73rpZseyrZ8","2252seNhipfTmjEBQ"]}