基于残差注意力多尺度关系网络的逻辑推理

doi:10.19678/j.issn.1000-3428.0064591

摘要/Abstract

摘要： 逻辑推理是感知视觉元素间规律和联系的能力，使计算机拥有类人的推理能力是一项重要的研究内容。在大量数据和深度模型的驱动下，现今人工智能已在图像处理等领域中取得超越人类水平的表现，但通过图像进行逻辑推理的能力还较落后。为解决面向逻辑推理的多尺度关系网络（MRNet）特征提取能力不足及泛化性较差的问题，提出一种改进的残差注意力多尺度关系网络（ResAMRNet）。在主干网络中，利用残差结构并结合跳跃连接与长跳跃连接，将浅层特征融入深层网络训练过程中，减少特征信息丢失，并提高模型特征提取能力。在推理模块中，将通道注意力机制与残差模块相融合检测每行图片间的关系特征，差异化各特征通道的重要程度，自适应学习注意力权重，提取关键特征。设计双池化高效通道注意力机制，结合全局最大池化进一步获取对象的特征信息，提高模型泛化性。在RAVEN和I-RAVEN数据集上的实验结果表明，ResAMRNet的分类准确率相比于MRNet分别提升了8.3和18.1个百分点，具有较强的逻辑推理能力。

关键词: 逻辑推理, 残差结构, 注意力机制, I-RAVEN数据集, 多尺度关系网络

Abstract: Logical reasoning is the ability to perceive patterns and connections between visual elements. Endowing computers with human-like reasoning ability is a critical area of research；state-of-the-art deep neural networks have achieved superhuman performance in image processing and other fields.However，the concept of logical reasoning through images requires further research.To address the problems of insufficient feature extraction and generalization of Multi-scale Relation Network（MRNet），an improved logical reasoning method，called Residual Attention Multi-scale Relation Network（ResAMRNet），is proposed. In the backbone network，shallow features are integrated into the deep network training process by utilizing residual structures and combining jump and long jump. This reduces the loss of feature information and improves the feature extraction capability of the model. In the reasoning module，the channel attention mechanism and residuals are combined to detect the relationship features between each image line.It can differentiate the significance of each feature channel，learn the attention weight adaptively，and extract key features.In this study，a Double-pooled Efficient Channel Attention（DECA） mechanism is proposed to combine global maximum pooling to further obtain feature information regarding objects and to improve generalization.Experimental results on representative logical reasoning datasets，Relational and Analogical Visual rEasoNing（RAVEN） and Improved RAVEN（I-RAVEN），show that the accuracy of the proposed method using these datasets is higher by 8.3 and 18.1 percentage points，respectively，than that of MRNet. Therefore，it demonstrates strong logical reasoning capabilities.

Key words: logical reasoning, residual structure, attention mechanism, I-RAVEN dataset, multi-scale relation network

中图分类号:

TP391

熊中敏, 曾旗, 卢鹏, 王振华, 郑宗生. 基于残差注意力多尺度关系网络的逻辑推理[J]. 计算机工程, 2023, 49(6): 227-233,241.

XIONG Zhongmin, ZENG Qi, LU Peng, WANG Zhenhua, ZHENG Zongsheng. Logical Reasoning Based on Residual Attention Multi-scale Relation Network[J]. Computer Engineering, 2023, 49(6): 227-233,241.

http://www.ecice06.com/CN/Y2023/V49/I6/227

图/表 12

20230615170245

20230615170249

20230615170252

20230615170256

20230615170259

20230615170305

20230615170308

20230615170312

20230615170315

20230615170319

20230615170324

20230615170327

参考文献

[1] CARPENTER P A,JUST M A,SHELL P.What one intelligence test measures:a theoretical account of the processing in the Raven progressive matrices test[J].Psychological Review,1990,97(3):404-431.
[2] RAVEN J C.Mental tests used in genetic studies:the performance of related individuals on tests mainly educative and mainly reproductive[D].London,UK:University of London,1936.
[3] RAVEN J.Raven progressive matrices[M].Berlin,Germany:Springer,2003.
[4] DOMINO G,DOMINO M L.Psychological testing:an introduction[M].2nd ed.Cambridge,UK:Cambridge University Press,2006.
[5] MAŁKINKI M,MANZIUK J.Deep learning methods for abstract visual reasoning:a survey on Raven's progressive matrices[EB/OL].[2022-03-17].https://arxiv.org/abs/2201.12382.
[6] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[7] SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[EB/OL].[2022-03-17].https://arxiv.org/abs/1409.1556.
[8] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2016:770-778.
[9] MAŁKINKI M,MANDZIUK J.A review of emerging research directions in abstract visual reasoning[J].Information Fusion,2023,91:713-736.
[10] ZHENG K,ZHA Z,WEI W.Abstract reasoning with distracting features[EB/OL].[2022-03-17].https://arxiv.org/abs/1912.00569.
[11] 丁硙,周枫,庙介璞,等.基于跨事件理论的新闻事件时序关系识别方法[J].计算机工程,2017,43(6):189-194.DING W,ZHOU F,MIAO J P,et al.Temporal relation recognition method for news events based on cross event theory[J].Computer Engineering,2017,43(6):189-194.(in Chinese)
[12] HERSCHE M,ZEQIRI M,BENINI L,et al.A neuro-vector-symbolic architecture for solving Raven's progressive matrices[EB/OL].[2022-03-17].https://arxiv.org/abs/2203.04571.
[13] ZHANG C,GAO F,JIA B X,et al.RAVEN:a dataset for relational and analogical visual reasoning[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2020:5312-5322.
[14] ZHANG C,JIA B,GAO F,et al.Learning perceptual inference by contrasting[C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems.New York,USA:ACM Press,2019:1075-1087.
[15] ZHUO T,KANKANHALLI M.Effective abstract reasoning with dual-contrast network[EB/OL].[2022-03-17].https://arxiv.org/abs/2205.13720.
[16] HU S,MA Y,LIU X,et al.Stratified rule-aware network for abstract visual reasoning[C]//Proceedings of 2021 AAAI Conference on Artificial Intelligence.Palo Alto,USA:AAAI Press,2021:1567-1574.
[17] BENNY Y,PEKAR N,WOLF L.Scale-localized abstract reasoning[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2021:12557-12565.
[18] ZAGORUYKO S,KOMODAKIS N.Learning to compare image patches via convolutional neural networks[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2015:4353-4361.
[19] 张婷.基于注意力机制的图像内容理解与视觉推理算法研究[D].成都:电子科技大学,2021.ZHANG T.Research on image content understanding and visual reasoning algorithm based on attention mechanism[D].Chengdu:University of Electronic Science and Technology of China,2021.(in Chinese)
[20] WANG Q,WU B,ZHU P,et al.ECA-Net:efficient channel attention for deep convolutional neural networks[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2020:11531-11539,2020.
[21] HU J,SHEN L,SUN G.Squeeze-and-excitation networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2018:7132-7141.
[22] 张学锋,李金晶.基于双注意力残差循环单幅图像去雨集成网络[J].软件学报,2021,32(10):3283-3292.ZHANG X F,LI J J.Single image de-raining using a recurrent dual-attention-residual ensemble network[J].Journal of Software,2021,32(10):3283-3292.(in Chinese)
[23] WOO S,PARK J,LEE J Y,et al.CBAM:convolutional block attention module[C]//Proceedings of European Conference on Computer Vision.Berlin,Germany:Springer,2018:3-19.
[24] KINGMA D P,BA J.Adam:a method for stochastic optimization[EB/OL].[2022-03-17].https://arxiv.org/abs/1412.6980.
[25] BARRETT D,HILL F,SANTORO A,et al.Measuring abstract reasoning in neural networks[C]//Proceedings of International Conference on Machine Learning.Washington D.C.,USA:IEEE Press,2018:511-520.
[26] GREFF K,SRIVASTAVA R K,KOUTNIK J,et al.LSTM:a search space Odyssey[J].IEEE Transactions on Neural Networks and Learning Systems,2017,28(10):2222-2232.

选择文件类型/文献管理软件名称

选择包含的内容