Key points are not available for this paper at this time.
In the current context of data explosion, online techniques that do not storing all data in memory are indispensable to routinely perform tasks principal component analysis (PCA). Recursive algorithms that update the with each new observation have been studied in various fields of research found wide applications in industrial monitoring, computer vision, , and latent semantic indexing, among others. This work provides for selecting an online PCA algorithm in practice. We present the main to online PCA, namely, perturbation techniques, incremental methods, stochastic optimization, and compare their statistical accuracy, time, and memory requirements using artificial and real data. to missing data and to functional data are discussed. All studied are available in the R package onlinePCA on CRAN.
Cardot et al. (Wed,) studied this question.