用于表示級(jí)特征融合與分類的相關(guān)熵融合極限學(xué)習(xí)機(jī)

吳超; 李雅倩; 張亞茹; 劉彬

doi:10.11999/JEIT190186

用于表示級(jí)特征融合與分類的相關(guān)熵融合極限學(xué)習(xí)機(jī)

doi: 10.11999/JEIT190186

吳超¹,
李雅倩^2, ,,
張亞茹¹,
劉彬¹

1.
燕山大學(xué)信息科學(xué)與工程學(xué)院秦皇島 066004
2.
燕山大學(xué)工業(yè)計(jì)算機(jī)控制工程河北省重點(diǎn)實(shí)驗(yàn)室秦皇島 066004

基金項(xiàng)目: 國(guó)家自然科學(xué)基金(51641609)

詳細(xì)信息

作者簡(jiǎn)介:
吳超：男，1990年生，博士生，研究方向?yàn)橛?jì)算機(jī)視覺(jué)

李雅倩：女，1982年生，副教授，研究方向?yàn)橛?jì)算機(jī)視覺(jué)

張亞茹：女，1995年生，博士生，研究方向?yàn)橛?jì)算機(jī)視覺(jué)

劉彬：男，1953年生，教授，研究方向?yàn)橛?jì)算機(jī)視覺(jué)

通訊作者:
李雅倩　yaqianli@126.com

中圖分類號(hào): TP391
計(jì)量
- 文章訪問(wèn)數(shù): 2750
- HTML全文瀏覽量: 1348
- PDF下載量: 112
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2019-03-27
- 修回日期: 2019-09-03
- 網(wǎng)絡(luò)出版日期: 2019-09-12
- 刊出日期: 2020-02-19

Correntropy-based Fusion Extreme Learning Machine forRepresentation Level Feature Fusion and Classification

Chao WU¹,
Yaqian LI^{2
, ,},
Yaru ZHANG¹,
Bin LIU¹

1.
School of Information Science and Engineering, Yanshan University, Qinhuangdao 066004, China
2.
Key Laboratory of Industrial Computer Control Engineering of Hebei Province, Yanshan University,Qinhuangdao 066004, China

Funds: The National Natural Science Foundation of China (51641609)

摘要

摘要:
在極限學(xué)習(xí)機(jī)(ELM)網(wǎng)絡(luò)結(jié)構(gòu)和訓(xùn)練模式的基礎(chǔ)上，該文提出了相關(guān)熵融合極限學(xué)習(xí)機(jī)(CF-ELM)。針對(duì)多數(shù)分類方法中表示級(jí)特征融合不充分的問(wèn)題，該文將核映射與系數(shù)加權(quán)相結(jié)合，提出了能夠有效融合表示級(jí)特征的融合極限學(xué)習(xí)機(jī)(F-ELM)。在此基礎(chǔ)上，用相關(guān)熵?fù)p失函數(shù)替代均方誤差(MSE)損失函數(shù)，推導(dǎo)出用于訓(xùn)練F-ELM各層權(quán)重矩陣的相關(guān)熵循環(huán)更新公式，以增強(qiáng)其分類能力與魯棒性。為了檢驗(yàn)方法的可行性，該文分別在數(shù)據(jù)庫(kù)Caltech 101, MSRC和15 Scene上進(jìn)行實(shí)驗(yàn)。實(shí)驗(yàn)結(jié)果證明，該文所提CF-ELM能夠在原有基礎(chǔ)上進(jìn)一步融合表示級(jí)特征，從而提高分類正確率。
- 極限學(xué)習(xí)機(jī) /
- 表示級(jí)特征融合 /
- 相關(guān)熵 /
- 分類
Abstract:
Based on the network structure and training methods of the Extreme Learning Machine (ELM), Correntropy-based Fusion Extreme Learning Machine (CF-ELM) is proposed. Considering the problem that the fusion of representation level features is insufficient in most classification methods, the kernel mapping and coefficient weighting are combined to propose a Fusion Extreme Learning Machine (F-ELM), which can effectively fuse the representation level features. On this basis, the Mean Square Error (MSE) loss function is replaced by the correntropy-based loss function. A correntropy-based cycle update formula for training the weight matrices of the F-ELM is derived to enhance classification ability and robustness. Extensive experiments are performed on Caltech 101, MSRC and 15 Scene datasets respectively. The experimental results show that CF-ELM can further fuse the representation level features to improve the classification accuracy.
- Extreme Learning Machine (ELM) /
- Representation level feature fusion /
- Correntropy /
- Classification

HTML全文

圖 1 融合極限學(xué)習(xí)機(jī)的網(wǎng)絡(luò)結(jié)構(gòu)

下載: 全尺寸圖片幻燈片

圖 2 CF-ELM中循環(huán)次數(shù)對(duì)Caltech 101與MSRC數(shù)據(jù)庫(kù)上正確率的影響曲線

下載: 全尺寸圖片幻燈片

圖 3 Caltech 101數(shù)據(jù)庫(kù)上2種組合的正確率曲面圖

下載: 全尺寸圖片幻燈片

圖 4 MSRC數(shù)據(jù)庫(kù)上3種組合的正確率曲面圖

下載: 全尺寸圖片幻燈片

圖 5 F-ELM與CF-ELM對(duì)兩種數(shù)據(jù)庫(kù)中具有嘈雜背景圖像的正確率

下載: 全尺寸圖片幻燈片

圖 6 CF-ELM中循環(huán)次數(shù)對(duì)15 Scene數(shù)據(jù)庫(kù)上正確率的影響曲線

下載: 全尺寸圖片幻燈片

圖 7 15 Scene數(shù)據(jù)庫(kù)上3種組合的正確率曲面圖

下載: 全尺寸圖片幻燈片

圖 8 F-ELM與CF-ELM對(duì)15 Scene的正確率

下載: 全尺寸圖片幻燈片

表 1 Caltech 101與MSRC的正確率與訓(xùn)練時(shí)間

組合方法	SVM			KELM			F-ELM			CF-ELM
組合方法	組合1	組合2	組合3	組合1	組合2	組合3	組合1	組合2	組合3	組合1	組合2	組合3
Caltech 101 (%)	72.04	77.93	–	79.43	78.84	–	80.31	80.19	–	80.59	83.65	–
訓(xùn)練時(shí)間(s)	862.16	540.30	–	860.37	538.56	–	861.76	539.77	–	902.13	569.80	–
MSRC (%)	90.26	88.57	94.13	91.42	90.42	91.74	91.74	90.58	93.49	90.95	92.06	95.76
訓(xùn)練時(shí)間(s)	79.59	17.11	17.11	79.51	17.01	17.01	79.54	17.04	17.04	80.68	18.20	18.20

下載: 導(dǎo)出CSV

表 2 Caltech 101的結(jié)果比較

方法	SPM^[9]	LLC^[17]	文獻(xiàn)[18]	SDCD+PHOW^[14]	文獻(xiàn)[19]	ScSPM+DVM^[20]	文獻(xiàn)[12]	文獻(xiàn)[8]	文獻(xiàn)[21]	文獻(xiàn)[16]	CF-ELM-組合2
字典維數(shù)	200	2048	1000	1024	–	–	400	1024	2048	600	400
正確率(%)	64.60	73.44	74.30	75.37	76.00	77.70	77.93	78.00	79.70	83.90	83.65

下載: 導(dǎo)出CSV

表 3 15 Scene的正確率與訓(xùn)練時(shí)間

組合方法	SVM			KELM			F-ELM			CF-ELM
組合方法	組合1	組合2	組合4	組合1	組合2	組合4	組合1	組合2	組合4	組合1	組合2	組合4
15 Scene(%)	74.34	77.92	86.46	72.00	80.73	83.53	77.20	82.06	84.33	79.00	83.06	87.76
訓(xùn)練時(shí)間(s)	347.44	106.25	106.25	347.12	106.11	106.11	347.37	106.41	106.41	367.50	120.10	120.10

下載: 導(dǎo)出CSV

表 4 15 Scene的結(jié)果比較

方法	SPM^[9]	LLC^[17]	SLC^[22]	LSVQ^[22]	文獻(xiàn)[23]	LGF^[15]	文獻(xiàn)[12]	MFS^[24]	文獻(xiàn)[16]	CF-ELM-組合4
字典維數(shù)	200	1000	1024	1024	–	1024	400	–	600	400
正確率(%)	81.10	81.73	81.89	83.08	85.70	85.80	86.46	87.10	90.10	87.76

下載: 導(dǎo)出CSV

參考文獻(xiàn)(24)

HUANG Guangbin, ZHU Qinyu, and SIEW C K. Extreme learning machine: Theory and applications[J]. Neurocomputing, 2006, 70(1/3): 489–501. doi: 10.1016/j.neucom.2005.12.126

HUANG Guangbin, ZHOU Hongming, DING Xiaojian, et al. Extreme learning machine for regression and multiclass classification[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B, 2012, 42(2): 513–529. doi: 10.1109/TSMCB.2011.2168604

KASUN L L C, ZHOU Hongming, HUANG Guangbin, et al. Representational learning with extreme learning machine for big data[J]. IEEE Intelligent Systems, 2013, 28(6): 31–34.

XING Hongjie and WANG Xinmei. Training extreme learning machine via regularized correntropy criterion[J]. Neural Computing and Applications, 2013, 23(7/8): 1977–1986. doi: 10.1007/s00521-012-1184-y

CHEN Liangjun, HONEINE P, QU Hua, et al. Correntropy-based robust multilayer extreme learning machines[J]. Pattern Recognition, 2018, 84: 357–370. doi: 10.1016/j.patcog.2018.07.011

LUO Xiong, SUN Jiankun, WANG Long, et al. Short-term wind speed forecasting via stacked extreme learning machine with generalized correntropy[J]. IEEE Transactions on Industrial Informatics, 2018, 14(11): 4963–4971. doi: 10.1109/TII.2018.2854549

HAN Honggui, WANG Lidan, and QIAO Junfei. Hierarchical extreme learning machine for feedforward neural network[J]. Neurocomputing, 2014, 128: 128–135. doi: 10.1016/j.neucom.2013.01.057

LI Qing, PENG Qiang, CHEN Junzhou, et al. Improving image classification accuracy with ELM and CSIFT[J]. Computing in Science & Engineering, 2019, 21(5): 26–34. doi: 10.1109/MCSE.2018.108164708

LAZEBNIK S, SCHMID C, and PONCE J. Beyond bags of features: Spatial pyramid matching for recognizing natural scene categories[C]. 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, New York, USA, 2006: 2169–2178. doi: 10.1109/CVPR.2006.68.

JÉGOU H, DOUZE M, SCHMID C, et al. Aggregating local descriptors into a compact image representation[C]. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, USA, 2010: 3304–3311. doi: 10.1109/CVPR.2010.5540039.

SáNCHEZ J, PERRONNIN F, MENSINK T, et al. Image classification with the fisher vector: Theory and practice[J]. International Journal of Computer Vision, 2013, 105(3): 222–245. doi: 10.1007/s11263-013-0636-x

李雅倩, 吳超, 李海濱, 等. 局部位置特征與全局輪廓特征相結(jié)合的圖像分類方法[J]. 電子學(xué)報(bào), 2018, 46(7): 1726–1731. doi: 10.3969/j.issn.0372-2112.2018.07.026

LI Yaqian, WU Chao, LI Haibin, et al. Image classification method combining local position feature with global contour feature[J]. Acta Electronica Sinica, 2018, 46(7): 1726–1731. doi: 10.3969/j.issn.0372-2112.2018.07.026

AHMED K T, IRTAZA A, and IQBAL M A. Fusion of local and global features for effective image extraction[J]. Applied Intelligence, 2017, 47(2): 526–543. doi: 10.1007/s10489-017-0916-1

MANSOURIAN L, ABDULLAH M T, ABDULLAH L N, et al. An effective fusion model for image retrieval[J]. Multimedia Tools and Applications, 2018, 77(13): 16131–16154. doi: 10.1007/s11042-017-5192-x

ZOU Jinyi, LI Wei, CHEN Chen, et al. Scene classification using local and global features with collaborative representation fusion[J]. Information Sciences, 2016, 348: 209–226. doi: 10.1016/j.ins.2016.02.021

KONIUSZ P, YAN Fei, GOSSELIN P H, et al. Higher-order occurrence pooling for bags-of-words: Visual concept detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(2): 313–326. doi: 10.1109/TPAMI.2016.2545667

WANG Jinjun, YANG Jianchao, YU Kai, et al. Locality-constrained linear coding for image classification[C]. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, USA, 2010: 3360–3367.

ZHU Qihai, WANG Zhezheng, MAO Xiaojiao, et al. Spatial locality-preserving feature coding for image classification[J]. Applied Intelligence, 2017, 47(1): 148–157. doi: 10.1007/s10489-016-0887-7

XIONG Wei, ZHANG Lefei, DU Bo, et al. Combining local and global: Rich and robust feature pooling for visual recognition[J]. Pattern Recognition, 2017, 62: 225–235. doi: 10.1016/j.patcog.2016.08.006

GUI Jie, LIU Tongliang, TAO Dacheng, et al. Representative vector machines: A unified framework for classical classifiers[J]. IEEE Transactions on Cybernetics, 2016, 46(8): 1877–1888. doi: 10.1109/TCYB.2015.2457234

GOH H, THOME N, CORD M, et al. Learning deep hierarchical visual feature coding[J]. IEEE Transactions on Neural Networks and Learning Systems, 2014, 25(12): 2212–2225. doi: 10.1109/TNNLS.2014.2307532

肖文華, 包衛(wèi)東, 陳立棟, 等. 一種用于圖像分類的語(yǔ)義增強(qiáng)線性編碼方法[J]. 電子與信息學(xué)報(bào), 2015, 37(4): 791–797. doi: 10.11999/JEIT140743

XIAO Wenhua, BAO Weidong, CHEN Lidong, et al. A semantic enhanced linear coding for image classification[J]. Journal of Electronics &Information Technology, 2015, 37(4): 791–797. doi: 10.11999/JEIT140743

LI Lijia, SU Hao, LIM Y, et al. Object bank: An object-level image representation for high-level visual recognition[J]. International Journal of Computer Vision, 2014, 107(1): 20–39. doi: 10.1007/s11263-013-0660-x

SONG Xinhang, JIANG Shuqiang, and HERRANZ L. Multi-scale multi-feature context modeling for scene recognition in the semantic manifold[J]. IEEE Transactions on Image Processing, 2017, 26(6): 2721–2735. doi: 10.1109/TIP.2017.2686017

相關(guān)文章

施引文獻(xiàn)

資源附件(0)

訪問(wèn)統(tǒng)計(jì)