Global Convergence of Adaptive Sensing for Principal Eigenvector Estimation
Researchers have developed a compressed version of Oja's algorithm for estimating the principal eigenvector of a data covariance matrix. This method requires only two adaptive measurements per sample, significantly reducing data acquisition needs. The analysis proves that the expected sine-squared error to the true eigenvector is bounded, establishing a theoretical limit for compressed eigenvector estimation and demonstrating its efficiency compared to non-adaptive schemes. AI