Cross-Validation for Unsupervised Learning
arXiv:0909.3052
Abstract
Cross-validation (CV) is a popular method for model-selection. Unfortunately, it is not immediately obvious how to apply CV to unsupervised or exploratory contexts. This thesis discusses some extensions of cross-validation to unsupervised learning, specifically focusing on the problem of choosing how many principal components to keep. We introduce the latent factor model, define an objective criterion, and show how CV can be used to estimate the intrinsic dimensionality of a data set. Through both simulation and theory, we demonstrate that cross-validation is a valuable tool for unsupervised learning.
Ph.D. thesis, Stanford University, 2009. Adviser: Art B. Owen. 165 pages
References in corpus (8)
- Level-Spacing Distributions and the Airy Kernel
- Limit theorems for sample eigenvalues in a generalized spiked population model
- A rate of convergence result for the largest eigenvalue of complex white Wishart matrices
- On the largest eigenvalue of Wishart matrices with identity covariance when n, p and p/n tend to infinity
- Are a set of microarrays independent of each other?
- Large deviations and stochastic calculus for large random matrices
- Minimax rank estimation for subspace tracking
- Random matrices: The distribution of the smallest singular values
Cited by in corpus (7)
- Clustering South African households based on their asset status using latent variable models
- The Optimal Hard Threshold for Singular Values is 4/sqrt(3)
- ScreeNOT: Exact MSE-Optimal Singular Value Thresholding in Correlated Noise
- Matrix Denoising with Partial Noise Statistics: Optimal Singular Value Shrinkage of Spiked F-Matrices
- Bi-cross-validation for factor analysis
- Fusing heterogeneous data sets
- Bi-cross validation for estimating spectral clustering hyper parameters