Fast PCA via UTV decomposition and application on EEG analysis.

Area of Science:

• Biomedical Signal Processing
• Machine Learning
• Data Analysis

Background:

• Principal Component Analysis (PCA), also known as the Karhunen-Loeve Transform (KLT), is effective for dimensionality reduction.
• Traditional PCA is computationally intensive and difficult to update for frequently changing biomedical data, such as electroencephalogram (EEG) signals, limiting its real-time application.

Purpose of the Study:

• To develop a computationally efficient method for approximating PCA that allows for easy data updating and downdating.
• To introduce a novel transform, fast PCA (fastPCA), suitable for real-time analysis of high-dimensional biomedical signals.

Main Methods:

• The proposed fastPCA is computed using UTV decomposition.
• UTV decomposition is a technique typically employed to approximate the rank-revealing properties of Singular Value Decomposition (SVD).