| 127 | 1 | 958 |
| 下载次数 | 被引频次 | 阅读次数 |
为了解决现有自编码器在噪声环境下由于L2损失和L2正则化导致图像重构效果不佳的问题，该文提出了一种新颖的基于平滑L1自编码器和稀疏特征学习的深度鲁棒图像重构方法(SL1AE-SFL),能够显著提升图像重构的性能和效果。首先，该文提出了平滑L1自编码器，采用L1损失函数和平滑sReLU激活函数，替代传统的L2损失函数和ReLU激活函数，从而提升了模型对噪声和异常数据的鲁棒性。其次，针对L2正则化难以有效抑制噪声特征影响的问题，该文提出了基于稀疏特征学习的鲁棒自编码器方法，包括SL1AE-SFL(L1)和SL1AE-SFL(L21),通过施加基于L1和L21范式的稀疏正则化约束，使得模型不被噪声特征所干扰、更加关注重要特征，从而显著增强了重构的准确性。最后，该文提出了一种基于近端梯度下降的高效优化算法，有效缓解了L1和L21稀疏正则化在训练过程中的低效问题。在多个数据集上的实验结果表明，SL1AE-SFL相较于其他方法在重构误差和聚类精度方面均表现出显著优势。此外，还分析了网络结构对重构质量的影响，消融实验表明，适当增加网络层数可以进一步提升图像重构效果，验证了深层网络在特征提取与图像重构方面的有效性。
Abstract：To address the issue of poor image reconstruction performance in noisy environments caused by L2 loss and L2 regularization in existing autoencoders, we propose a novel deep robust image reconstruction method based on a Smoothed L1 Autoencoder and Sparse Feature Learning(SL1AE-SFL),which significantly enhances reconstruction quality and performance. Firstly, we introduce the Smoothed L1 Autoencoder, which replaces the traditional L2 loss function and ReLU activation function with the L1 loss function and smoothed sReLU activation function, thereby improving the model's robustness against noises and outliers. Secondly, to tackle the challenge of L2 regularization failing to effectively suppress the influence of noisy features, we propose a robust autoencoder approach based on sparse feature learning, including SL1AE-SFL(L1) and SL1AE-SFL(L21). By imposing sparse regularization constraints based on L1 and L21 norms, the model becomes less affected by noisy features and focuses more on essential features, thereby significantly improving reconstruction accuracy. Finally, we propose an efficient optimization algorithm based on proximal gradient descent, which effectively alleviates the training inefficiency of L1-norm and L21-norm based sparse regularizations. Experimental results on multiple datasets demonstrate that SL1AE-SFL outperforms other methods in terms of reconstruction error and clustering accuracy. Moreover, we analyze the impact of network architecture on reconstruction quality, and ablation studies show that appropriately increasing the number of network layers can further enhance image reconstruction performance, validating the effectiveness of deep networks in feature extraction and image reconstruction.
[1] 吴美君，杨新，潘超凡，等.自编码器结合持续学习：现状、挑战与展望[J].计算机学报，2025,48(2):317-357.
[2] 周思聪，向峰，李红军，等.基于自编码器GAN数据增强的工业小目标缺陷检测[J].现代制造工程，2025(2):101-108.
[3] 李善莲，安佳敏，刘朝贤，等.基于自编码器和PCA的滚筒烘丝机多块过程监测方法[J].轻工学报，2023,38(6):110-117.
[4] 唐肝翌，卢桂馥，王勇，等.基于L1-范数和弹性网约束的鲁棒稀疏块PCA[J].南京师大学报：自然科学版，2022,45(4):102-109.
[5] NIE F,TIAN L,HUANG H,et al.Non-greedy L21-norm maximization for principal component analysis[J].IEEE Transactions on Image Processing,2021,30:5277-5286.
[6] 陈平，刘珂菡，梁正友，等.基于稀疏协同相关熵的鲁棒主成分分析[J/OL].计算机科学.http://kns.cnki.net/kcms/detail/50.1075.TP.20241226.1100.009.html.
[7] HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[8] BALIN M F,ABID A,ZOU J.Concrete autoencoders:differentiable feature selection and reconstruction[C]//International conference on machine learning.Long Beach:PMLR,2019:444-453.
[9] LEE W,PARK S,KIM T.Denoising graph autoencoder for missing human joints reconstruction[J].IEEE Access,2024,12:57381-57389.
[10] PUCCI R,MARTINEL N.CE-VAE:capsule enhanced variational autoencoder for underwater image enhancement[C]//Proceedings of the winter conference on applications of computer vision.Tucson:WACV,2025:2113-2123.
[11] LEE Y,KWON H,PARK F.Neighborhood reconstructing autoencoders[C]//Advances in neural information processing systems.California:NeurIPS,2021:536-546.
[12] LEE Y,YOON S,SON M J,et al.Regularized autoencoders for isometric representation learning[C]//International conference on learning representations.Washington DC:ICLR,2022.
[13] LIM J,KIM J,LEE Y,et al.Graph geometry-preserving autoencoders[C]//Proceedings of the 41st international conference on machine learning.Honolulu:PMLR,2024:29795-29815.
[14] WANG Q,LIU M,FENG W,et al.Reconstruction weighting principal component analysis with fusion contrastive learning[C]//Proceedings of the 33rd international joint conference on artificial intelligence.Jeju Island:IJCAI,2024:5091-5098.
[15] HE K,CHEN X,XIE S,et al.Masked autoencoders are scalable vision learners[C]//Proceedings of the 2022 IEEE/CVF conference on computer vision and pattern recognition.New Orleans:IEEE,2022:16000-16009.
[16] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Advances in neural information processing systems.Lake Tahoe:NIPS,2012:1097-1105.
[17] BJORCK J,WEINBERGER K Q,GOMES C.Understanding decoupled and early weight decay[C]//Proceedings of the 35th AAAI conference on artificial intelligence.Palo Alto:AAAI,2021:6777-6785.
[18] ZHANG J,YAO J,YAN L,et al.Sparse multi-relational graphconvolutional network for multi-type object trajectory prediction[C]//Proceedings of the 33rd international joint conference on artificial intelligence.Jeju Island:IJCAI,2024:1697-1705.
[19] 李家强，杨广乐，徐必勇，等.基于分式平滑l0范数的穿墙雷达稀疏成像算法[J].中国电子科学研究院学报，2022,17(12):1147-1153.
[20] 陈慧，陈适，郭银婷，等.基于正则自编码器及Optuna寻优的异常用电数据清洗研究[J].电力需求侧管理，2023,25(5):53-58.
[21] TIAN Y,ZHANG Y,ZHANG H.Recent advances in stochasticgradient descent in deep learning[J].Mathematics,2023,11(3):682.
[22] XIAO G,ZHANG S,BAI Z J.Scaled proximal gradient methods for sparse optimization problems[J].Journal of Scientific Computing,2024,98(1):2.
[23] KINGMA D P,BA J L.Adam:a method for stochastic optimization[C]//Proceedings of the 3rd international conference on learning representations.San Diego:ICLR,2015.
[24] SAMARIA F S,HARTER A C.Parameterisation of a stochastic model for human face identification[C]//Proceedings of the 1994 IEEE workshop on applications of computer vision.Sarasota:IEEE,1994:138-142.
[25] MARTINEZ A M,KAK A C.PCA versus LDA[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2001,23(2):228-233.
[26] GEORGHIADES A S,BELHUMEUR P N,KRIEGMAN D J.From few to many:illumination cone models for face recognition under variable lighting and pose[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,23(6):643-660.
[27] NEFIAN A V,HAYES M H.An embedded HMM-based approach for face detection and recognition[C]//Proceedings of the 1999 IEEE international conference on acoustics,speech,and signal processing.Phoenix:IEEE,1999:3553-3556.
基本信息:
DOI：10.20165/j.cnki.ISSN1673-629X.2025.0138
中图分类号:TP391.41;TP18
引用信息:
[1]张俊洋,明镝.基于自编码器和稀疏学习的深度鲁棒图像重构[J].计算机技术与发展,2025,35(10):43-52.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0138.
基金信息:
国家自然科学基金(62441607); 重庆市自然科学基金面上项目(CSTB2024NSCQ-MSX0341); 重庆市教委科学技术研究项目(KJQN202301142); 重庆理工大学科研启动基金资助项目(2022ZDZ026)
2025-06-09
2025-06-09
2025-06-09