Filters








127,282 Hits in 3.4 sec

A simple neural network module for relational reasoning [article]

Adam Santoro, David Raposo, David G.T. Barrett, Mateusz Malinowski, Razvan Pascanu, Peter Battaglia, Timothy Lillicrap
2017 arXiv   pre-print
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.  ...  Acknowledgments We would like to thank Murray Shanahan, Ari Morcos, Scott Reed, Daan Wierstra, Alex Lerchner, and many others on the DeepMind team, for critical feedback and discussions.  ... 
arXiv:1706.01427v1 fatcat:t7tycni3rndfdl4ofhux4vr2a4

Working Memory Networks: Augmenting Memory Networks with a Relational Reasoning Module

Juan Pavez, Héctor Allende, Héctor Allende-Cid
2018 Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)  
During the last years, there has been a lot of interest in achieving some kind of complex reasoning using deep neural networks.  ...  To solve these issues, we introduce the Working Memory Network, a MemNN architecture with a novel working memory storage and reasoning module.  ...  To do that we pass the output through a simple neural network f t .  ... 
doi:10.18653/v1/p18-1092 dblp:conf/acl/PavezAA18 fatcat:ecc2tayjljgcrhygazxajpu244

Working Memory Networks: Augmenting Memory Networks with a Relational Reasoning Module [article]

Juan Pavez, Héctor Allende, Héctor Allende-Cid
2018 arXiv   pre-print
During the last years, there has been a lot of interest in achieving some kind of complex reasoning using deep neural networks.  ...  To solve these issues, we introduce the Working Memory Network, a MemNN architecture with a novel working memory storage and reasoning module.  ...  To do that we pass the output through a simple neural network f t .  ... 
arXiv:1805.09354v1 fatcat:qzqlac5tzfbsnbj6o5vq6mtqey

Dilated DenseNets for Relational Reasoning [article]

Antreas Antoniou, Agnieszka Słowik, Elliot J. Crowley, Amos Storkey
2018 arXiv   pre-print
Despite their impressive performance in many tasks, deep neural networks often struggle at relational reasoning.  ...  In this extended abstract, we show that a DenseNet incorporating dilated convolutions excels at relational reasoning on the Sort-of-CLEVR dataset, allowing us to forgo this relational module and its associated  ...  A simple neural network module for relational reasoning.  ... 
arXiv:1811.00410v1 fatcat:njsdwionurdyvgsfmmooyya2uu

Neural Module Networks [article]

Jacob Andreas, Marcus Rohrbach, Trevor Darrell, Dan Klein
2017 arXiv   pre-print
We describe a procedure for constructing and learning *neural module networks*, which compose collections of jointly-trained neural "modules" into deep networks for question answering.  ...  We evaluate our approach on two challenging datasets for visual question answering, achieving state-of-the-art results on both the VQA natural image dataset and a new dataset of complex questions about  ...  Acknowledgments The authors are grateful to Lisa Anne Hendricks, Eric Tzeng, and Russell Stewart for useful conversations, and to Nvidia for a hardware grant.  ... 
arXiv:1511.02799v4 fatcat:4lrnuh4tafdevcl3ixcmxqufvi

What Can Neural Networks Reason About? [article]

Keyulu Xu, Jingling Li, Mozhi Zhang, Simon S. Du, Ken-ichi Kawarabayashi, Stefanie Jegelka
2020 arXiv   pre-print
Neural networks have succeeded in many reasoning tasks.  ...  Empirically, these tasks require specialized network structures, e.g., Graph Neural Networks (GNNs) perform well on many such tasks, but less structured networks fail.  ...  PRELIMINARIES We begin by introducing notations and summarizing common neural networks for reasoning tasks. Let S denote the universe, i.e., a configuration/set of objects to reason about.  ... 
arXiv:1905.13211v4 fatcat:qvuc7uncovfjhdbgzaajxgkyfu

Toward Accurate Visual Reasoning With Dual-Path Neural Module Networks

Ke Su, Hang Su, Jianguo Li, Jun Zhu
2020 Frontiers in Robotics and AI  
The recently proposed neural module network (Andreas et al., 2016b), which assembles the model with a few primitive modules, is capable of performing a spatial or arithmetical reasoning over the input  ...  To address these issues, we propose a novel method of Dual-Path Neural Module Network which can implement complex visual reasoning by forming a more flexible layout regularized by the pairwise loss.  ...  of composing a new deep network with neural modules for each given input.  ... 
doi:10.3389/frobt.2020.00109 pmid:33501276 pmcid:PMC7805672 fatcat:4p6abhbp4vhlninrbgdy2bg6xu

RelNet: End-to-End Modeling of Entities & Relations [article]

Trapit Bansal, Arvind Neelakantan, Andrew McCallum
2017 arXiv   pre-print
We introduce RelNet: a new model for relational reasoning.  ...  RelNet is a memory augmented neural network which models entities as abstract memory slots and is equipped with an additional relational memory which models relations between all memory pairs.  ...  They demonstrated that simple neural network modules are not as effective at relational reasoning and their proposed module is similar to our model.  ... 
arXiv:1706.07179v2 fatcat:gv4tx6ecdja2fiem6fztl25oau

Learning to Compose Neural Networks for Question Answering [article]

Jacob Andreas, Marcus Rohrbach, Trevor Darrell, Dan Klein
2016 arXiv   pre-print
The model uses natural language strings to automatically assemble neural networks from a collection of composable modules.  ...  Parameters for these modules are learned jointly with network-assembly parameters via reinforcement learning, with only (world, question, answer) triples as supervision.  ...  Acknowledgments JA is supported by a National Science Foundation Graduate Fellowship. MR is supported by a fellowship within the FIT weltweit-Program of the German Academic Exchange Service (DAAD).  ... 
arXiv:1601.01705v4 fatcat:kt2skogwirerpny37poacfr2e4

Learning to Compose Neural Networks for Question Answering

Jacob Andreas, Marcus Rohrbach, Trevor Darrell, Dan Klein
2016 Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies  
The model uses natural language strings to automatically assemble neural networks from a collection of composable modules.  ...  Our approach, which we term a dynamic neural module network, achieves state-of-theart results on benchmark datasets in both visual and structured domains.  ...  Acknowledgments JA is supported by a National Science Foundation Graduate Fellowship. MR is supported by a fellowship within the FIT weltweit-Program of the German Academic Exchange Service (DAAD).  ... 
doi:10.18653/v1/n16-1181 dblp:conf/naacl/AndreasRDK16 fatcat:n5hnyyiuafaohl24qeybjpgzbi

FigureNet: A Deep Learning model for Question-Answering on Scientific Plots [article]

Revanth Reddy, Rahul Ramesh, Ameet Deshpande, Mitesh M. Khapra
2019 arXiv   pre-print
Our approach outperforms the state-of-the-art Relation Networks baseline by approximately 7% on this dataset, with a training time that is over an order of magnitude lesser.  ...  In this work, we describe a deep learning model that addresses the reasoning task of question-answering on categorical plots.  ...  Relation Networks Relation networks(RN) were introduced by [10] as a simple yet powerful neural module for relational reasoning.  ... 
arXiv:1806.04655v2 fatcat:okpc5v6krzfmjoxuas7b3nnvvi

Predicting the Popularity of Online Videos via Deep Neural Networks [article]

Yue Mao, Yi Shen, Gang Qin, Longjun Cai
2017 arXiv   pre-print
In this paper, we propose a general prediction model using the multi-task learning (MTL) module and the relation network (RN) module, where MTL can reduce over-fitting and RN can model the relations of  ...  Predicting the popularity of online videos is important for video streaming content providers. This is a challenging problem because of the following two reasons.  ...  An RN is a simple and effective module for the case when input "objects" (usually hidden units of a neural network) have relations which each other.  ... 
arXiv:1711.10718v2 fatcat:3zxjlnijkbforll4spg3vjkn4y

KANDINSKYPatterns – An experimental exploration environment for Pattern Analysis and Machine Intelligence [article]

Andreas Holzinger, Anna Saranti, Heimo Mueller
2021 arXiv   pre-print
There is still a significant gap between machine-level pattern recognition and human-level concept learning.  ...  Humans can learn under uncertainty from only a few examples and generalize these concepts to solve new problems.  ...  ACKNOWLEDGEMENTS This work has received funding by the Austrian Science Fund (FWF), Project: P-32554 "A reference model for explainable Artificial Intelligence in the medical domain".  ... 
arXiv:2103.00519v1 fatcat:d57pwgzf4vhmpaa5hqwm7ls5zq

Graph Neural Networks with Generated Parameters for Relation Extraction [article]

Hao Zhu, Yankai Lin, Zhiyuan Liu, Jie Fu, Tat-seng Chua, Maosong Sun
2019 arXiv   pre-print
Among existing models, graph neural networks (GNNs) is one of the most effective approaches for multi-hop relational reasoning.  ...  In this paper, we propose to generate the parameters of graph neural networks (GP-GNNs) according to natural language sentences, which enables GNNs to process relational reasoning on unstructured text  ...  For example, Santoro et al. (2017) propose a simple neural network to reason the relationship of objects in a picture, Xu et al. (2017) build up a scene graph according to an image, and (Kipf et al  ... 
arXiv:1902.00756v1 fatcat:lpnbujhf4nd5hdu7iltiif7jwu

Unifying Neural Learning and Symbolic Reasoning for Spinal Medical Report Generation [article]

Zhongyi Han, Benzheng Wei, Yilong Yin, Shuo Li
2020 arXiv   pre-print
Concretely, we design an adversarial graph network that interpolates a symbolic graph reasoning module into a generative adversarial network through embedding prior domain knowledge, achieving semantic  ...  In this paper, we propose the neural-symbolic learning (NSL) framework that performs human-like learning by unifying deep neural learning and symbolic logical reasoning for the spinal medical report generation  ...  In this paper, we propose a new graph reasoning module for capturing the relations between spinal structures and improving the high-level semantic representation.  ... 
arXiv:2004.13577v1 fatcat:oh5aka5zr5be3ipd7qqnyhikzy
« Previous Showing results 1 — 15 out of 127,282 results