ML//unsupervised learning

Methods that find structure in data without labels: directions of variation, groups, and low-dimensional representations.

Methods that find structure in data without labels: directions of variation, groups, and low-dimensional representations.

PCA keeps the directions of largest variance. K-means groups points around centroids. spectral clustering first represents connectivity between samples, then groups.

Their mathematical foundations live under Math: the covariance matrix, projection and spectral coordinates.