方法对比

	奇异谱分析	独立成分分析（ICA）	核主成分分析
领域	时间序列	机器学习	机器学习
方法族	Process / pipeline	Latent structure	Latent structure
起源年份	1986	1994	1998
提出者	David Broomhead	Comon, P.	Schölkopf, B.; Smola, A. J.; Müller, K.-R.
类型	Dimension reduction and trend extraction	Blind source separation / latent-structure decomposition	Nonlinear dimensionality reduction via kernel trick
开创性文献	Broomhead, D. S., & King, G. P. (1986). Extracting qualitative dynamics from experimental data. Physica D: Nonlinear Phenomena, 20(2–3), 217–236.	Comon, P. (1994). Independent component analysis, a new concept? Signal Processing, 36(3), 287–314.	Schölkopf, B., Smola, A. J., & Müller, K.-R. (1998). Nonlinear component analysis as a kernel eigenvalue problem. Neural Computation, 10(5), 1299–1319.
别名	SSA, SVD-based decomposition	ICA, blind source separation, BSS, FastICA	KPCA, kernel PCA, nonlinear PCA via kernel trick, kernel eigenvalue decomposition
相关	3	3	5
摘要	Singular Spectrum Analysis (SSA) is a nonparametric method for time-series decomposition and forecasting based on singular value decomposition (SVD) of a time-lagged embedding matrix. Introduced by Broomhead and King (1986) and developed further by Vautard, Yiou, and Ghil (1992), SSA decomposes time series into trend, oscillatory, and noise components without assuming any underlying model. It is particularly effective for short, noisy non-stationary signals where parametric approaches fail.	Independent Component Analysis (ICA) is a computational method for separating a multivariate signal into additive, statistically independent subcomponents. Formalized by Pierre Comon in 1994, ICA became the foundational framework for blind source separation and is widely applied in neuroimaging (fMRI, EEG), speech processing, and biomedical signal analysis.	Kernel Principal Component Analysis (Kernel PCA) is a nonlinear dimensionality-reduction method introduced by Bernhard Schölkopf, Alexander Smola, and Klaus-Robert Müller in 1997–1998. It extends classical linear PCA to curved, non-linear data manifolds by implicitly mapping input data into a high-dimensional feature space via a kernel function, then performing standard PCA in that space — all without ever computing the mapping explicitly.