应用于非精确图匹配的改进DF模型

doi:10.3778/j.issn.1673-9418.2011060

计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (6): 1383-1389.DOI: 10.3778/j.issn.1673-9418.2011060

应用于非精确图匹配的改进DF模型

李智杰(), 伊志林, 李昌华, 张颉

西安建筑科技大学信息与控制工程学院,西安 710055

收稿日期:2020-11-23 修回日期:2021-02-26 出版日期:2022-06-01 发布日期:2021-03-25
通讯作者: + E-mail: lizhijie@xauat.edu.cn
作者简介:李智杰（1980—）,男,河南人,博士,副教授,主要研究方向为模式识别、数字建筑等。
伊志林（1990—）,男,山东人,硕士研究生,主要研究方向为模式识别、深度学习等。
李昌华（1963—）,男,宁夏人,博士,教授,博士生导师,主要研究方向为图形图像处理、模式识别、数字建筑等。
张颉（1989—）,男,陕西人,博士研究生,主要研究方向为数字建筑、模式识别等。
基金资助:
国家自然科学基金(61373112);国家自然科学基金(51878536);陕西省自然科学基金(2020JQ-687);陕西省住房城乡建设科技计划项目(2020-K09)

Improved DF Model Applied to Inexact Graph Matching

LI Zhijie(), YI Zhilin, LI Changhua, ZHANG Jie

School of Information and Control Engineering, Xi’an University of Architecture and Technology, Xi’an 710055, China

Received:2020-11-23 Revised:2021-02-26 Online:2022-06-01 Published:2021-03-25
About author:LI Zhijie, born in 1980, Ph.D., associate professor. His research interests include pattern recognition, digital architecture, etc.
YI Zhilin, born in 1990, M.S. candidate. His research interests include pattern recognition, deep learning, etc.
LI Changhua, born in 1963, Ph.D., professor, Ph.D. supervisor. His research interests include graph and image processing, pattern recognition, digital architecture, etc.
ZHANG Jie, born in 1989, Ph.D. candidate. His research interests include digital architecture, pattern recognition, etc.
Supported by:
National Natural Science Foundation of China(61373112);National Natural Science Foundation of China(51878536);Natural Science Foundation of Shaanxi Province(2020JQ-687);Science and Technology Project of Housing and Urban-Rural Construction of Shaanxi Province(2020-K09)

摘要/Abstract

摘要：

针对传统深度森林算法提取的特征不够完整,以及采取的等权决策机制对分类结果易产生差异性等问题,提出一种应用于非精确图匹配的改进DF模型（IDF）。首先,在挖掘特征子集的过程中,采取融合移动窗口和随机移动窗口的方式。在移动窗口扫描样本的同时,随机捕获一个与移动扫描窗口相同大小的特征子集,两者构成新的特征子集,从而将新特征子集作为级联森林模块的输入。其次,在级联森林的迭代过程中,计算当前森林的决策结果所占权重,并与上一级森林进行对比,采用Min的权值策略规则赋值给当前森林,逐次迭代直至结果满足模型所设定的阈值。最后,在MUTAG、PTC、COX2等数据集上进行了训练和测试。实验结果表明,相比于传统深度森林算法,IDF充分考虑了图的结构特征,能够有效增强样本的拟合优度及多样性,降低了级联模块中各子树的决策差异及模型的复杂度,有效提升了模型的分类识别率。

关键词: 非精确图匹配, 深度森林, 决策树, 加权

Abstract:

Aiming at the problems that the features extracted by the traditional deep forest algorithm are not complete, and the equal-power decision mechanism is easy to produce differences in the classification results, an improved DeepForest (IDF) model applied to inexact graph matching is proposed. Firstly, in the process of mining feature subsets, the methods of fusing moving windows and random moving windows are adopted. While the moving window scans the sample, a same size feature subset of the moving scanning window is randomly captured, and these form a new feature subset, which is used as the input of the cascade module. Secondly, in the iterative process of the cascading forest, the weight of decision result in the current forest is calculated. Compared with the upper level forest, the weight value is assigned to the current forest by the strategy rule of Min, and iteration is continued until the result meets the given threshold value by the model. Finally, training and testing are conducted on datasets such as MUTAG, PTC and COX2. The experimental results show that, compared with traditional deep forest algorithm, IDF fully considers the structural characteristics of the graph, and can effectively enhance the diversity of samples and the goodness of fit, and reduce the decision-making difference and the complexity of the model. It efectively improves the classification and recognition rate of the model.

Key words: inexact graph matching, deep forest, decision tree, weighted

中图分类号:

TP391.4

李智杰, 伊志林, 李昌华, 张颉. 应用于非精确图匹配的改进DF模型[J]. 计算机科学与探索, 2022, 16(6): 1383-1389.

LI Zhijie, YI Zhilin, LI Changhua, ZHANG Jie. Improved DF Model Applied to Inexact Graph Matching[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(6): 1383-1389.

图/表 9

图1 复合多粒度扫描

Fig.1 Compound multi-grained scanning

图2 单层级联森林概率分布

Fig.2 Probability distribution of single-level cascade forest

图3 模型结构关系

Fig.3 Relationship of model structure

表1 数据集信息

Table 1 Information of dataset

数据集	特征数	类别数	备注
MUTAG	188	2	该化合物是否对细菌均有诱变作用
PTC	344	2	该化合物是否对老鼠有致癌行为
COX2	467	2	该分类表示在体外对人体重组酶是否有抑制作用

图4 复合多粒度深度森林准确率

Fig.4 Accuracy of compound multi-grained deep forest

图5 加权级联深度森林准确率

Fig.5 Accuracy of weighted cascade deep forest

图6 三个数据集上两种算法的准确率

Fig.6 Accuracy of two algorithms on three datasets

表2 深度模型层数

Table 2 The number of layers in depth model

实验模型	深度模型层数
实验模型	MUTAG	PTC	COX2
SAE	14	15	22
DF	8	7	12
IDF	6	6	10

表3 数据集的平均准确率

Table 3 Average accuracy of dataset

实验模型	平均准确率/%
实验模型	MUTAG	PTC	COX2
IDF	94.5	73.7	94.1
DF	91.7	71.3	90.7
CMDF	92.2	72.1	91.8
WCDF	93.1	72.5	91.8
BCNNS	89.4	73.4	92.3

参考文献 18

[1]	严骏驰, 杨小康. 计算机视觉中图匹配研究进展: 从二图匹配迈向多图匹配[J]. 控制理论与应用, 2018, 35(12): 1715-1724.
	YAN J C, YANG X K. Recent advance on graph matching in computer vision: from two-graph matching to multi-graph matching[J]. Control Theory & Applications, 2018, 35(12): 1715-1724.
[2]	李智杰, 李昌华, 刘欣, 等. 融合拓扑特征和领域特征的非精确图匹配算法[J]. 计算机应用与软件, 2015, 32(10): 164-167.
	LI Z J, LI C H, LIU X, et al. Inexact graph matching algo-rithm integrating topological features and domain features[J]. Computer Applications and Software, 2015, 32(10): 164-167.
[3]	ZHANG J K, QIAN K. Graph matching using conformal module[J]. EURASIP Journal on Image and Video Processing, 2019: 26.
[4]	刘国庆, 卢桂馥, 周胜, 等. 非负低秩图嵌入算法[J]. 计算机科学与探索, 2020, 14(3): 502-512.
	LIU G Q, LU G F, ZHOU S, et al. Non-negative low rank graph embedding algorithm[J]. Journal of Frontiers of Com-puter Science and Technology, 2020, 14(3): 502-512.
[5]	许文, 宋文爱, 富丽贞, 等. 面向大规模图数据的分布式子图匹配算法[J]. 计算机科学, 2019, 46(4): 28-35.
	XU W, SONG W A, FU L Z, et al. Distributed subgraph matching algorithm for large scale graph data[J]. Computer Science, 2019, 46(4): 28-35.
[6]	ZHOU Z H, FENG J. Deep forest: towards an alternative to deep neural networks[J]. arXiv:1702.08835, 2017.
[7]	张西宁, 郭清林, 刘书语. 深度学习技术及其故障诊断应用分析与展望[J]. 西安交通大学学报, 2020, 54(12): 1-13.
	ZHANG X N, GUO Q L, LIU S Y. Analysis and prospect of deep learning technology and its fault diagnosis applica-tion[J]. Journal of Xi’an Jiaotong University, 2020, 54(12): 1-13.
[8]	佟彤, 罗森林, 潘丽敏, 等. 基于深度森林的量表数据挖掘方法[J]. 电子设计工程, 2020, 28(13): 88-91.
	TONG T, LUO S L, PAN L M, et al. Deep forest based inven-tory data mining method[J]. Electronic Design Engineering, 2020, 28(13): 88-91.
[9]	葛绍林, 叶剑, 何明祥. 基于深度森林的用户购买行为预测模型[J]. 计算机科学, 2019, 46(9): 190-194.
	GE S L, YE J, HE M X. Prediction model of user purchase behavior based on deep forest[J]. Computer Science, 2019, 46(9): 190-194.
[10]	陈寅栋, 李朝锋, 桑庆兵. 卷积神经网络结合深度森林的无参考图像质量评价[J]. 激光与光电子学进展, 2019, 56(11): 123-129.
	CHEN Y D, LI C F, SANG Q B. Quality assessment without reference images based on convolution neural network and deep forest[J]. Laser & Optoelectronics Progress, 2019, 56(11): 123-129.
[11]	余星达, 陈文杰, 王鼎, 等. 非接触式身份识别的深度学习算法[J]. 西安交通大学学报, 2019, 53(4): 122-127.
	YU X D, CHEN W J, WANG D, et al. A deep learning algori-thm for contactless human identification[J]. Journal of Xi’an Jiaotong University, 2019, 53(4): 122-127.
[12]	李昌华, 崔李扬, 李智杰. 用于非精确图匹配的改进GCN模型[J]. 计算机科学与探索, 2020, 14(8): 1397-1408.
	LI C H, CUI L Y, LI Z J. Improved GCN model for inexact graph matching[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(8): 1397-1408.
[13]	乔安, 毛力, 孙俊. 基于改进深度森林的小目标检测算法[J]. 传感器与微系统, 2020, 39(5): 125-128.
	QIAO A, MAO L, SUN J. Small target detection algorithm based on improved deep forest[J]. Transducer and Micro-system Technologies, 2020, 39(5): 125-128.
[14]	ZHANG Q, XU Y. Block-based selection random forest for texture classification using multi-fractal spectrum feature[J]. Neural Computing and Applications, 2016, 27(3): 593-602. DOI URL
[15]	UTKIN L V. An imprecise deep forest for classification[J]. Expert Systems with Applications, 2020, 141: 112978. DOI URL
[16]	宫振华, 王嘉宁, 苏翀. 一种加权的深度森林算法[J]. 计算机应用与软件, 2019, 36(2): 274-278.
	GONG Z H, WANG J N, SU C. A weighted deep forest algorithm[J]. Computer Applications and Software, 2019, 36(2): 274-278.
[17]	UTKIN L V, KOVALEV M, MELDO A A. A deep forest classifier with weights of class probability distribution subsets[J]. Knowledge-Based Systems, 2019, 173: 15-27. DOI URL
[18]	尹儒, 门昌骞, 王文剑. 一种模型决策森林算法[J]. 计算机科学与探索, 2020, 14(1): 108-116.
	YIN R, MEN C Q, WANG W J. Model decision forest algo-rithm[J]. Journal of Frontiers of Computer Science and Tech-nology, 2020, 14(1): 108-116.

应用于非精确图匹配的改进DF模型

Improved DF Model Applied to Inexact Graph Matching

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 18

相关文章 15

编辑推荐

Metrics

[1]	鱼先锋, 耿生玲. 模糊智能决策树模型与应用研究[J]. 计算机科学与探索, 2022, 16(3): 703-712.
[2]	夏笑秋, 陈松灿. 改进的二视图随机森林[J]. 计算机科学与探索, 2022, 16(1): 144-152.
[3]	武晓栋, 刘敬浩, 金杰, 毛思平. 基于DT及PCA的DNN入侵检测模型[J]. 计算机科学与探索, 2021, 15(8): 1450-1458.
[4]	范瑞东, 侯臣平. 鲁棒自加权的多视图子空间聚类[J]. 计算机科学与探索, 2021, 15(6): 1062-1073.
[5]	梁凌, 邓赵红, 王士同. 兼顾显隐信息与特征加权的多视角模糊聚类[J]. 计算机科学与探索, 2021, 15(6): 1092-1102.
[6]	张炜, 邓赵红, 王士同. 基于核诱导的不完整多视角聚类[J]. 计算机科学与探索, 2021, 15(2): 284-293.
[7]	薛红艳, 钱雪忠, 周世兵. 超簇加权的集成聚类算法[J]. 计算机科学与探索, 2021, 15(12): 2362-2373.
[8]	李昌华，崔李扬，李智杰. 用于非精确图匹配的改进GCN模型[J]. 计算机科学与探索, 2020, 14(8): 1397-1408.
[9]	罗浩，王彦捷，牛明航，邱存月，张利. 动态区间的加权模糊聚类算法[J]. 计算机科学与探索, 2020, 14(7): 1142-1153.
[10]	杜师帅，邱天，李灵巧，胡锦泉，郑安兵，冯艳春，胡昌勤，杨辉华. 多层梯度提升树在药品鉴别中的应用[J]. 计算机科学与探索, 2020, 14(2): 260-273.
[11]	胡健，徐锴滨，毛伊敏. 基于加权网格和信息熵的并行密度聚类算法[J]. 计算机科学与探索, 2020, 14(12): 2094-2107.
[12]	尹儒，门昌骞，王文剑. 一种模型决策森林算法[J]. 计算机科学与探索, 2020, 14(1): 108-116.
[13]	王小玉，韩昌林，胡鑫豪. 加权特征融合的密集连接网络人脸识别算法[J]. 计算机科学与探索, 2019, 13(7): 1195-1205.
[14]	魏明桦，郑金贵. 自适应目标与内容匹配的层级图像分割算法[J]. 计算机科学与探索, 2019, 13(4): 681-692.
[15]	陈家益，战荫伟，曹会英，吴兴达，李小飞. 修剪中值检测的自适应加权中值滤波算法[J]. 计算机科学与探索, 2019, 13(3): 505-513.