High Dimensional Semiparametric Scale-Invariant Principal Component Analysis.

Area of Science:

• Statistics
• Machine Learning
• Data Analysis

Background:

• Principal Component Analysis (PCA) is a widely used dimensionality reduction technique.
• Existing methods like sparse PCA have limitations in robustness and interpretability.

Purpose of the Study:

• To introduce Copula Component Analysis (COCA), a new high-dimensional semiparametric PCA method.
• To demonstrate COCA's advantages over traditional PCA and sparse PCA.

Main Methods:

• Developed a semiparametric model assuming multivariate Gaussian distributions after monotone transformations.
• Utilized Copula Component Analysis (COCA) for high-dimensional data.

Main Results:

• COCA estimators achieve fast estimation rates and feature selection consistency.
• COCA demonstrates robustness to modeling assumptions and outliers.
• COCA is scale-invariant, leading to more interpretable results.
• Experiments show COCA outperforms sparse PCA on synthetic and real-world data.

Conclusions:

• COCA offers a robust and interpretable alternative for high-dimensional semiparametric PCA.
• The method provides theoretical guarantees for estimation and feature selection.
• COCA shows superior performance compared to existing techniques.