Online Principal Component Analysis in High Dimension: Which Algorithm to Choose?
From MaRDI portal
Publication:6086546
Abstract: In the current context of data explosion, online techniques that do not require storing all data in memory are indispensable to routinely perform tasks like principal component analysis (PCA). Recursive algorithms that update the PCA with each new observation have been studied in various fields of research and found wide applications in industrial monitoring, computer vision, astronomy, and latent semantic indexing, among others. This work provides guidance for selecting an online PCA algorithm in practice. We present the main approaches to online PCA, namely, perturbation techniques, incremental methods, and stochastic optimization, and compare their statistical accuracy, computation time, and memory requirements using artificial and real data. Extensions to missing data and to functional data are discussed. All studied algorithms are available in the R package onlinePCA on CRAN.
Cites work
- A Stable and Efficient Algorithm for the Rank-One Modification of the Symmetric Eigenproblem
- Acceleration of Stochastic Approximation by Averaging
- Adaptive algorithms for first principal eigenvector computation
- Almost sure convergence of stochastic gradient processes with matrix step sizes
- ARPACK Users' Guide
- Finding structure with randomness: probabilistic algorithms for constructing approximate matrix decompositions
- Functional data analysis.
- Global convergence of Oja's PCA learning algorithm with a non-zero-approaching adaptive learning rate
- scientific article; zbMATH DE number 3642533 (Why is no real title available?)
- scientific article; zbMATH DE number 3497161 (Why is no real title available?)
- scientific article; zbMATH DE number 3533576 (Why is no real title available?)
- scientific article; zbMATH DE number 976356 (Why is no real title available?)
- scientific article; zbMATH DE number 2045498 (Why is no real title available?)
- scientific article; zbMATH DE number 6159604 (Why is no real title available?)
- Low-dimensional tracking of association structures in categorical data
- Low-rank incremental methods for computing dominant singular subspaces
- Method of stochastic approximation in the determination of the largest eigenvalue of the mathematical expectation of random matrices
- Multiple imputation in principal component analysis
- Multivariate data analysis: The French way
- On stochastic approximation of the eigenvectors and eigenvalues of the expectation of a random matrix
- On Updating Problems in Latent Semantic Indexing
- Online principal components analysis
- Principal component analysis.
- Randomized online PCA algorithms with regret bounds that are logarithmic in the dimension
- Rank-one modification of the symmetric eigenproblem
- Scikit-learn: machine learning in Python
- Sequential Karhunen-Loève basis extraction and its application to images
- Some Modified Matrix Eigenvalue Problems
Cited in
(21)- Recursive principal component analysis for model order reduction with application in nonlinear Bayesian filtering
- Chunk-wise regularised PCA-based imputation of missing data
- Widening the scope of an eigenvector stochastic approximation process and application to streaming PCA and related methods
- Unsupervised streaming anomaly detection for instrumented infrastructure
- PLS for Big Data: a unified parallel algorithm for regularised group PLS
- Quantifying deviations from separability in space-time functional processes
- Incremental modelling for compositional data streams
- Online sparse sliced inverse regression for high-dimensional streaming data
- Minimum cost‐compression risk in principal component analysis
- Stochastic Gauss-Newton algorithms for online PCA
- Sparse online principal component analysis for parameter estimation in factor model
- Online inference in high-dimensional generalized linear models with streaming data
- Statistical Analysis of Random Objects Via Metric Measure Laplacians
- An extended Oja process for streaming canonical analysis
- Online inference in high-dimensional regression with streaming clustered data
- Incremental singular value decomposition for some numerical aspects of multiblock redundancy analysis
- Online kernel sliced inverse regression
- An orthonormalization-free and inversion-free algorithm for online estimation of principal eigenspace
- Online landmark replacement for out-of-sample dimensionality reduction methods
- Online debiased lasso estimation and inference for heterogenous updating regressions
- Convergence analysis of Oja's iteration for solving online PCA with nonzero-mean samples
This page was built for publication: Online Principal Component Analysis in High Dimension: Which Algorithm to Choose?
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6086546)
