[Verse 1] Every matrix holds a secret, locked inside its rows and columns A equals U times sigma times V transpose, the golden formula Rectangular or square, it matters not at all SVD reveals what eigenvalues could never show [Chorus] U sigma V transpose, the trinity of truth Left and right singular vectors, sigma holds the proof Orthogonal foundations, diagonal between The X-ray vision breaking matrices clean U sigma V transpose, decomposition's crown Strip away complexity, bare the skeleton down [Verse 2] U is m by m, the left side guardian Sigma diagonal, descending like a staircase V transpose on the right, n by n and orthogonal Count the nonzero sigmas, that's your matrix rank revealed [Chorus] U sigma V transpose, the trinity of truth Left and right singular vectors, sigma holds the proof Orthogonal foundations, diagonal between The X-ray vision breaking matrices clean U sigma V transpose, decomposition's crown Strip away complexity, bare the skeleton down [Bridge] Truncate to the top k, Eckart-Young will guarantee The best approximation that you'll ever see Sigma max divided by sigma minimum Condition number whispers perturbation's hymn [Verse 3] Column space and null space, row space and its left companion Four fundamental subspaces dancing in formation Compress your data, analyze components Recommender systems, pseudoinverses potent [Chorus] U sigma V transpose, the trinity of truth Left and right singular vectors, sigma holds the proof Orthogonal foundations, diagonal between The X-ray vision breaking matrices clean U sigma V transpose, decomposition's crown Strip away complexity, bare the skeleton down [Outro] When eigendecomposition fails you SVD will always sail through The strongest bones remain standing Mathematical understanding
← 8 The Spectral Theorem | 10 Dual Spaces and Tensor Products (Introduction) →