面向不平衡圖像數(shù)據(jù)的對抗自編碼器過采樣算法
doi: 10.11999/JEIT240330
-
鄭州大學計算機與人工智能學院 鄭州 450001
基金項目: 國家重點研發(fā)計劃 (2023YFC2206404)
Adversarial Autoencoders Oversampling Algorithm for Imbalanced Image Data
-
School of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, China
Funds: The National Key Research and Development Project (2023YFC2206404)
-
摘要: 許多適用于低維數(shù)據(jù)的傳統(tǒng)不平衡學習算法在圖像數(shù)據(jù)上的效果并不理想?；谏蓪咕W(wǎng)絡(GAN)的過采樣算法雖然可以生成高質量圖像，但在類不平衡情況下容易產(chǎn)生模式崩潰問題?；谧跃幋a器(AE)的過采樣算法容易訓練，但生成的圖像質量較低。為進一步提高過采樣算法在不平衡圖像中生成樣本的質量和訓練的穩(wěn)定性，該文基于生成對抗網(wǎng)絡和自編碼器的思想提出一種融合自編碼器和生成對抗網(wǎng)絡的過采樣算法(BAEGAN)。首先在自編碼器中引入一個條件嵌入層，使用預訓練的條件自編碼器初始化GAN以穩(wěn)定模型訓練；然后改進判別器的輸出結構，引入一種融合焦點損失和梯度懲罰的損失函數(shù)以減輕類不平衡的影響；最后從潛在向量的分布映射中使用合成少數(shù)類過采樣技術(SMOTE)來生成高質量的圖像。在4個圖像數(shù)據(jù)集上的實驗結果表明該算法在生成圖像質量和過采樣后的分類性能上優(yōu)于具有輔助分類器的條件生成對抗網(wǎng)絡(ACGAN)、平衡生成對抗網(wǎng)絡 (BAGAN)等過采樣算法，能有效解決圖像數(shù)據(jù)中的類不平衡問題。
-
關鍵詞:
- 不平衡圖像數(shù)據(jù) /
- 過采樣 /
- 生成對抗網(wǎng)絡 /
- 對抗自編碼器 /
- 合成少數(shù)類過采樣技術
Abstract: Many traditional imbalanced learning algorithms suitable for low-dimensional data do not perform well on image data. Although the oversampling algorithm based on Generative Adversarial Networks (GAN) can generate high-quality images, it is prone to mode collapse in the case of class imbalance. Oversampling algorithms based on AutoEncoders (AE) are easy to train, but the generated images are of lower quality. In order to improve the quality of samples generated by the oversampling algorithm in imbalanced images and the stability of training, a Balanced oversampling method with AutoEncoders and Generative Adversarial Networks (BAEGAN) is proposed in this paper, which is based on the idea of GAN and AE. First, a conditional embedding layer is introduced in the Autoencoder, and the pre-trained conditional Autoencoder is used to initialize the GAN to stabilize the model training; then the output structure of the discriminator is improved, and a loss function that combines Focal Loss and gradient penalty is proposed to alleviate the impact of class imbalance; and finally the Synthetic Minority Oversampling TEchnique (SMOTE) is used to generate high-quality images from the distribution map of latent vectors. Experimental results on four image data sets show that the proposed algorithm is superior to oversampling methods such as Auxiliary Classifier Generative Adversarial Networks (ACGAN) and BAlancing Generative Adversarial Networks (BAGAN) in terms of image quality and classification performance after oversampling and can effectively solve the class imbalance problem in image data. -
1 BAEGAN算法描述
輸入：從不平衡的訓練集$X$中劃分一批數(shù)據(jù)$B = \{ {b_1},{b_2},\cdots,$
${b_{|X|/m}}\} $；批量大小$m$；類別數(shù)量$n$；預先設定的模型超參數(shù)；
先驗分布$p({\boldsymbol{z}})$；輸出：平衡后的數(shù)據(jù)集${X_{\text}}$ (1) (a) 初始化所有網(wǎng)絡參數(shù)(編碼器${\theta _E}$、解碼器${\theta _{{\text{De}}}}$、生成器
${\theta _G}$、判別器${\theta _D}$)，預訓練條件自編碼器：(2) WHILE預訓練輪數(shù) DO： (3) FOR 從$B$中選取一組數(shù)據(jù)$({\boldsymbol{x}},{\boldsymbol{c}})$ DO： (4) 將數(shù)據(jù)${\boldsymbol{x}}$送入編碼器$E$，獲得${\boldsymbol{z}}$； (5) 將${\boldsymbol{z}}$和${\boldsymbol{c}}$輸入嵌入層，獲得${{\boldsymbol{z}}_{\text{c}}}$； (6) 將${{\boldsymbol{z}}_{\text{c}}}$送入解碼器${\text{De}}$，獲得重構圖像$\hat {\boldsymbol{x}}$； (7) 由式(2)計算損失，更新${\theta _E}$和${\theta _{{\text{De}}}}$。 (8) END (9) END (10) (b) 預訓練的條件自編碼器初始化${\theta _G}$和${\theta _{{\text{De}}}}$，訓練模型： (11) WHILE 模型未收斂或未達到訓練輪數(shù) DO： (12) FOR 從$B$中選取一組數(shù)據(jù)$({\boldsymbol{x}},{\boldsymbol{c}})$ DO： (13) 將數(shù)據(jù)${\boldsymbol{x}}$送入編碼器$E$中，獲得${\boldsymbol{z}}$； (14) 將${\boldsymbol{z}}$和${\boldsymbol{c}}$輸入嵌入層中，獲得${{\boldsymbol{z}}_{\text{c}}}$； (15) 將${{\boldsymbol{z}}_{\text{c}}}$送入解碼器${\text{De}}$，獲得重構圖像$\hat {\boldsymbol{x}}$； (16) 根據(jù)式(2)計算損失，更新${\theta _E}$和${\theta _{{\text{De}}}}$。 (17) 將${\boldsymbol{x}}$送入$G$，獲得${{\boldsymbol{z}}_{{\text{fake}}}}$ ，從$p({\boldsymbol{z}})$中獲得${{\boldsymbol{z}}_{{\text{real}}}}$； (18) 將${{\boldsymbol{z}}_{{\text{fake}}}}$和${{\boldsymbol{z}}_{{\text{real}}}}$輸入判別器$D$，由式(4)計算判別器損失，
更新${\theta _D}$；(19) ${{\boldsymbol{z}}_{{\text{fake}}}}$送入$D$，由式(5)計算生成器損失，更新${\theta _G}$； (20) END (21) END (22) (c) 生成樣本，平衡數(shù)據(jù)集： (23) WHILE 選取少數(shù)類${{c}}$中的所有樣本$({{\boldsymbol{x}}_{\text{c}}},{\boldsymbol{c}})$ ，直至所有少數(shù)
類選取完畢DO：(24) 將數(shù)據(jù)${{\boldsymbol{x}}_{\mathrm{c}}}$送入$E$中，獲得潛在向量${\boldsymbol{z}}$； (25) 將${\boldsymbol{z}}$和${\boldsymbol{c}}$送入SMOTE中，獲得平衡后的潛在向量${{\boldsymbol{z}}^{\text}}$和類
標簽${{\boldsymbol{c}}^{\text}}$；(26) 將${{\boldsymbol{z}}^{\text}}$和${{\boldsymbol{c}}^{\text}}$輸入嵌入層中，獲得嵌入條件的向量$ {\boldsymbol{z}}_{\text{c}}^{\text} $； (27) 將$ {\boldsymbol{z}}_{\text{c}}^{\text} $送入解碼器${\text{De}}$，獲得平衡后屬于類${\text{c}}$的樣本集； (28) END (29) 獲得平衡數(shù)據(jù)集${X_{\text}}$。 
下載: 導出CSV
表 1 網(wǎng)絡結構設置
層數(shù) 卷積核數(shù)量 卷積核大小 步長 填充 判別器或編碼器 1 64 4 2 1 2 128 4 2 1 3 256 4 2 1 4 512 4 2 1 生成器或解碼器 1 512 4 1 0 2 256 4 2 1 3 128 4 2 1 4 64 4 2 1 5 圖像通道數(shù) 4 2 1
下載: 導出CSV
表 3 不同過采樣算法在各不平衡數(shù)據(jù)集上的分類性能
算法 MNIST FMNIST SVHN CIFAR-10 ACSA F1 GM ACSA F1 GM ACSA F1 GM ACSA F1 GM CGAN[12] 0.8792 0.8544 0.9057 0.6528 0.6362 0.7263 0.7259 0.6908 0.7936 0.3319 0.3088 0.5302 ACGAN[13] 0.9212 0.9123 0.9492 0.8144 0.7895 0.8606 0.7720 0.7403 0.8239 0.4006 0.3410 0.5918 BAGAN[16] 0.9306 0.9277 0.9598 0.8148 0.8093 0.8931 0.8023 0.7775 0.8677 0.4338 0.4025 0.6373 DeepSMOTE[11] 0.9609 0.9603 0.9780 0.8363 0.8327 0.9061 0.8094 0.7873 0.8739 0.4538 0.4335 0.6530 BAEGAN 0.9807 0.9715 0.9842 0.8799 0.8156 0.9133 0.8357 0.7769 0.8942 0.5443 0.5254 0.7301
下載: 導出CSV
表 4 在CIFAR-10上消融實驗分類結果
算法 ACSA F1 GM BAEGAN-AE 0.4226 0.3946 0.5802 BAEGAN-L 0.3584 0.3142 0.4098 BAEGAN-S 0.2732 0.2233 0.3083 BAEGAN 0.5443 0.5254 0.7301
下載: 導出CSV
-
[1] FAN Xi, GUO Xin, CHEN Qi, et al. Data augmentation of credit default swap transactions based on a sequence GAN[J]. Information Processing & Management, 2022, 59(3): 102889. doi: 10.1016/j.ipm.2022.102889. [2] 劉俠, 呂志偉, 李博, 等. 基于多尺度殘差雙域注意力網(wǎng)絡的乳腺動態(tài)對比度增強磁共振成像腫瘤分割方法[J]. 電子與信息學報, 2023, 45(5): 1774–1785. doi: 10.11999/JEIT220362.LIU Xia, Lü Zhiwei, LI Bo, et al. Segmentation algorithm of breast tumor in dynamic contrast-enhanced magnetic resonance imaging based on network with multi-scale residuals and dual-domain attention[J]. Journal of Electronics & Information Technology, 2023, 45(5): 1774–1785. doi: 10.11999/JEIT220362. [3] 尹梓諾, 馬海龍, 胡濤. 基于聯(lián)合注意力機制和一維卷積神經(jīng)網(wǎng)絡-雙向長短期記憶網(wǎng)絡模型的流量異常檢測方法[J]. 電子與信息學報, 2023, 45(10): 3719–3728. doi: 10.11999/JEIT220959.YIN Zinuo, MA Hailong, and HU Tao. A traffic anomaly detection method based on the joint model of attention mechanism and one-dimensional convolutional neural network-bidirectional long short term memory[J]. Journal of Electronics & Information Technology, 2023, 45(10): 3719–3728. doi: 10.11999/JEIT220959. [4] FERNáNDEZ A, GARCíA S, GALAR M, et al. Learning From Imbalanced Data Sets[M]. Cham: Springer, 2018: 327–349. doi: 10.1007/978-3-319-98074-4. [5] HUANG Zhan’ao, SANG Yongsheng, SUN Yanan, et al. A neural network learning algorithm for highly imbalanced data classification[J]. Information Sciences, 2022, 612: 496–513. doi: 10.1016/j.ins.2022.08.074. [6] FU Saiji, YU Xiaotong, and TIAN Yingjie. Cost sensitive ν-support vector machine with LINEX loss[J]. Information Processing & Management, 2022, 59(2): 102809. doi: 10.1016/j.ipm.2021.102809. [7] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]. The IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2999–3007. doi: 10.1109/ICCV.2017.324. [8] LI Buyu, LIU Yu, and WANG Xiaogang. Gradient harmonized single-stage detector[C]. The 33rd AAAI Conference on Artificial Intelligence, Honolulu, USA, 2019: 8577–8584. doi: 10.1609/aaai.v33i01.33018577. [9] MICHELUCCI U. An introduction to autoencoders[J]. arXiv preprint arXiv: 2201.03898, 2022. doi: 10.48550/arXiv.2201.03898. [10] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]. The 27th International Conference on Neural Information Processing Systems, Montreal, Canada, 2014: 2672–2680. [11] DABLAIN D, KRAWCZYK B, and CHAWLA N V. DeepSMOTE: Fusing deep learning and SMOTE for imbalanced data[J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(9): 6390–6404. doi: 10.1109/TNNLS.2021.3136503. [12] MIRZA M and OSINDERO S. Conditional generative adversarial nets[J]. arXiv preprint arXiv: 1411.1784, 2014. doi: 10.48550/arXiv.1411.1784. [13] ODENA A, OLAH C, and SHLENS J. Conditional image synthesis with auxiliary classifier GANs[C]. The 34th International Conference on Machine Learning, Sydney, Australia, 2017: 2642–2651. [14] GULRAJANI I, AHMED F, ARJOVSKY M, et al. Improved training of wasserstein GANs[C]. The 31st International Conference on Neural Information Processing Systems, Long Beach, USA, 2017: 5769–5779. [15] CHAWLA N V, BOWYER K W, HALL L O, et al. SMOTE: Synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321–357. doi: 10.1613/jair.953. [16] MARIANI G, SCHEIDEGGER F, ISTRATE R, et al. BAGAN: Data augmentation with balancing GAN[J]. arXiv preprint arXiv: 1803.09655, 2018. doi: 10.48550/arXiv.1803.09655. [17] HUANG Gaofeng and JAFARI A H. Enhanced balancing GAN: Minority-class image generation[J]. Neural Computing and Applications, 2023, 35(7): 5145–5154. doi: 10.1007/s00521-021-06163-8. [18] BAO Jianmin, CHEN Dong, WEN Fang, et al. CVAE-GAN: Fine-grained image generation through asymmetric training[C]. The IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2764–2773. doi: 10.1109/ICCV.2017.299. [19] MAKHZANI A, SHLENS J, JAITLY N, et al. Adversarial autoencoders[J]. arXiv preprint arXiv: 1511.05644, 2015. doi: 10.48550/arXiv.1511.05644. [20] CUI Yin, JIA Menglin, LIN T Y, et al. Class-balanced loss based on effective number of samples[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, USA, 2019: 9260–9269. doi: 10.1109/CVPR.2019.00949. [21] DOWSON D C and LANDAU B V. The Fréchet distance between multivariate normal distributions[J]. Journal of Multivariate Analysis, 1982, 12(3): 450–455. doi: 10.1016/0047-259X(82)90077-X. [22] HUANG Chen, LI Yining, LOY C C, et al. Learning deep representation for imbalanced classification[C]. The IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 5375–5384. doi: 10.1109/CVPR.2016.580. [23] KUBAT M and MATWIN S. Addressing the curse of imbalanced training sets: One-sided selection[C]. 14th International Conference on Machine Learning, Nashville, USA, 1997: 179–186. [24] HRIPCSAK G and ROTHSCHILD A S. Agreement, the F-measure, and reliability in information retrieval[J]. Journal of the American Medical Informatics Association, 2005, 12(3): 296–298. doi: 10.1197/jamia.M1733. [25] SOKOLOVA M and LAPALME G. A systematic analysis of performance measures for classification tasks[J]. Information Processing & Management, 2009, 45(4): 427–437. doi: 10.1016/j.ipm.2009.03.002. [26] RADFORD A, METZ L, and CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[C]. 4th International Conference on Learning Representations, San Juan, Puerto Rico, 2016. -
圖(9) / 表(6)
計量
- 文章訪問數(shù): 175
- HTML全文瀏覽量: 128
- PDF下載量: 40
- 被引次數(shù): 0
下載:
