classical-mlunsupervisedrepresentation
PCA and dimensionality reduction
PCA, t-SNE, UMAP, autoencoders — finding low-dimensional structure in high-dimensional data.
Уровни глубины
L0Intro~2ч
Projects 3D points onto 2D; reads a PCA scatterplot.
L1Basics~10ч
Derives PCA via SVD / eigendecomposition of covariance; picks k by explained variance.
L2Working~15ч
Uses t-SNE / UMAP for visualisation; knows when linear PCA fails; kernel PCA.
L3Advanced~25ч
Probabilistic PCA, factor analysis, ICA; connects to autoencoders.
L4Research~50ч
Manifold learning theory, contrastive representations.