Information-theoretic approaches
Abstract
An entirely different but extremely relevant approach to dimensionality reduction can be taken using a different criterion, namely quantifying the information content of the features involved, within themselves or in relation to others. It turns out that Shannon's definition of information yields surprisingly interesting reductions. This chapter discusses five major variations of this idea, including comparisons using the concept of mutual information previously used in statistics and machine learning.
Publication Title
Dimensionality Reduction in Data Science
Recommended Citation
Garzon, M., Mainali, S., & Jana, K. (2022). Information-theoretic approaches. Dimensionality Reduction in Data Science, 127-144. https://doi.org/10.1007/978-3-031-05371-9_6
COinS
