Conventional statistical approaches
Abstract
The objective of dimensionality reduction is to retain key properties of the given data to solve a problem with fewer features in a lower dimensional space. Statistical methods aim to preserve characteristic parameters such as mean, variance, and covariance of features in the population, as estimated from the dataset. Methods include Principal Component Analysis (PCA) and its variants, Independent component analysis and Discriminant Analysis. Linear algebra methods offer other approaches, including Singular value Decomposition (SVD) and Nonnegative Matrix Factorization (NMF).
Publication Title
Dimensionality Reduction in Data Science
Recommended Citation
Yang, C., Garzon, M., & Deng, L. (2022). Conventional statistical approaches. Dimensionality Reduction in Data Science, 79-95. https://doi.org/10.1007/978-3-031-05371-9_4
