It is the process of reducing the number of variables under consideration by obtaining a set of principal variables.
Card 1 of 259. Answer: It is the process of reducing the number of variables under consideration by obtaining a set of principal variables.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
Dimensionality reduction means representing data with fewer variables while keeping as much of its useful structure as possible. It works either by selecting a subset of the original columns or by building a smaller set of new variables from combinations of them, as principal component analysis does.
The motivation is the curse of dimensionality. High-dimensional space is mostly empty: 10 features at 10 levels each make 10¹⁰ cells, and as dimensions grow, distances between random points become nearly equal, hurting nearest-neighbor methods and clustering. Real data usually sits near a lower-dimensional surface: MNIST digit images have 784 pixels, yet PCA on the raw pixels keeps 95% of their variance with about 150 components.
It removes noise and collinearity before modeling, compresses data and lets you plot it in two or three dimensions. Interviewers ask about the costs: extracted features lose their meaning, and an unsupervised method like PCA can drop a low-variance direction that predicts the target. In t-SNE and UMAP plots, cluster sizes and gaps are not reliable.
