在本篇論文研究中，我們依據兩個睡眠多項生理功能檢查（PSG）資料庫的腦電圖（EEG）信號來進行睡眠階段的分類。在該資料庫中，腦電波的特徵是通過散射變換來擷取。由於腦電波訊號的紀錄來自多個頭皮區塊，我們使用了多視角最小平方支持向量機模型（Multi-view LS-SVM）來進行分析，該模型同時考慮了多個來源EEG信號之間的相互作用。由於需要對5個睡眠階段（包括清醒狀態，快速動眼期與三個非快速動眼期階段）進行分類，我們將二元分類SVM推廣到多個類別的分類，採用一對多（OVA）編碼在每個類與所有其他類之間進行二元分類決策。為了識別資料中的離群值，我們使用敏感度曲線分析檢測並觀察偵測到的數據點對分類模型所產生的影響。在這裡我們使用留一法則交叉驗證（LOSOCV）來評估演算法的效能。此外，我們對多視角與單視角的最小平方支持向量機模型進行分類成果比較，並採用合成少數類樣本的增量技術（SMOTE）處理資料類别不平衡的問題。根據實際數據分析的結果，我們得出結論，在睡眠階段識別中，多視角最小平方支持向量機模型的分類效果優於單視角模型。

In this study, we focus on identifying the sleep stages based on the Electroencephalography (EEG) signals from two standard overnight Polysomnography (PSG) databases, where the features of the brain waves are extracted by the scattering transform. Because the EEG signals were collected from the multiple channels, we use the Multi-view Least Squares Support Vector Machine model (MV-LSSVM), which considers the interaction of the EEG signals between channel and channel simultaneously. We first generalize the classical SVM from a binary classification to a multi-class classification since the 5 sleep stages, including Awake, REM, N1, N2, and N3, need to be classified. The one-versus-all (OVA) encoding is employed to make the binary decisions between each class and all the other classes. To identify the outlying data points in the dataset, we apply the sensitivity curve and observe the effect of the detected data points on the prediction model. The leave-one-subject-out cross-validation (LOSOCV) is applied to evaluate the performance of the studied algorithms. Besides, we make a capability comparison between the multi-view LS-SVM and single view LS-SVM, and employ the Synthetic Minority Over-sampling Technique (SMOTE) to deal with the imbalanced classification problem. Finally, based on the results of the real data examples, we conclude that the multi-view classification on LS-SVM outperforms the single view classification in the sleep stage identification.

[1] J. Andén and S. Mallat. Deep scattering spectrum.IEEE Transactions on SignalProcessing, 62(16):4114–4128, 2014.
[2] Bernhard E. Boser, Isabelle M. Guyon, and Vladimir N. Vapnik. A training algo-rithm for optimal margin classifiers. InProceedings of the Fifth Annual Workshopon Computational Learning Theory, pages 144–152, New York, NY, USA, 1992.Association for Computing Machinery.
[3] Nitesh V. Chawla, Kevin W. Bowyer, Lawrence O. Hall, and W. PhilipKegelmeyer. Smote: Synthetic minority over-sampling technique.J. Artif. Int.Res., 16(1):321–357, 2002.
[4] Andreas Christmann and Ingo Steinwart. On robustness properties of convexrisk minimization methods for pattern recognition.Journal of Machine LearningResearch, 5:1007–1034, 2004.
[5] Corinna Cortes and Vladimir N. Vapnik. Support-vector networks.MachineLearning, 20(3):273–297, 1995.
[6] N. Cristianini and J. Shawe-Taylor.An Introduction to Support Vector Machinesand Other Kernel-based Learning Methods.Cambridge: Cambridge UniversityPress., 2000.
[7] Frank R. Hampel. The influence curve and its role in robust estimation.Journalof the American Statistical Association, 69(346):383–393, 1974.
[8] Frank R. Hampel, Elvezio M. Ronchetti, Peter J. Rousseeuw, and Werner A.Stahel. Robust statistics: the approach based on influence functions. 1986.
[9] Lynn Houthuys, Rocco Langone, and Johan A.K. Suykens. Multi-view leastsquares support vector machines classification.Neurocomputing, 282:78–88, 2018.
[10] Gi-Ren Liu, Yu-Lun Lo, John Malik, Yuan-Chung Sheu, and Hau-Tieng Wu.Diffuse to fuse eeg spectra–intrinsic geometry of sleep dynamics for classification.Biomedical Signal Processing and Control, 55:101576, 2020.
[11] Gi Ren Liu, Caroline Lustenberger, Yu Lun Lo, Wen Te Liu, Yuan Chung Sheu,and Hau Tieng Wu. Save muscle information-unfiltered eeg signal helps distinguishsleep stages.Sensors, 20(7), 2020.
[12] J. Mercer. Functions of positive and negative type, and their connection with thetheory of integral equations.Proceedings of the Royal Society of London, Series A:Containing Papers of a Mathematical and Physical Character, 209:415–446, 1909.
[13] Bernhard Scholkopf and Alexander J. Smola.Learning with Kernels: SupportVector Machines, Regularization, Optimization, and Beyond. MIT Press, 2001.
[14] Shiliang Sun. A survey of multi-view machine learning.Neural Computing andApplications, 23(7):2031–2038, 2013.
[15] J.A.K. Suykens and J. Vandewalle. Least squares support vector machine classi-fiers.Neural Processing Letters, 9(3):293–300, 1999.
[16] John W. Tukey. A survey of sampling from contaminated distributions. 1959.
[17] John W. Tukey.Exploratory Data Analysis. Reading, Mass: Addison-Wesley Pub.Co., 1977.
[18] V. N. Vapnik and A. Ya. Lerner. Recognition of patterns with help of generalizedportraits.Avtomat. i Telemekh., 24(6):774–780, 1963.
[19] Vladimir N. Vapnik.The Nature of Statistical Learning Theory. Springer-Verlag,Berlin, Heidelberg, 1995.
[20] Vladimir N. Vapnik.Statistical Learning Theory. Wiley-Interscience, 1998.
[21] Grace Wahba.Support Vector Machines, Reproducing Kernel Hilbert Spaces, andRandomized GACV. MIT Press, Cambridge, MA, USA, 1999.
[22] Jing Zhao, Xijiong Xie, Xin Xu, and Shiliang Sun. Multi-view learning overview:Recent progress and new challenges.Information Fusion, 38:43–54, 2017