机器学习的几何观点-LAMDAGroup.pptVIP

A Geometric Perspective on Machine Learning;Machine Learning: the problem;Manifold Learning: geometric perspective;Manifold Learning: the challenges ;Manifold Learning: current solution;Simplicity;Simplicity;Simplicity is relative;Manifold-based Dimensionality Reduction;A Good Mapping Function ;Locality Preserving Projections;Locality Preserving Projections;Locality Preserving Projections;Locality Preserving Projections;Manifold of Face Images;Manifold of Handwritten Digits;Learning target: Training Examples: Linear Regression Model ;Generalization Error;Gauss-Markov Theorem;Gauss-Markov Theorem;Experimental Design Methods;Manifold Regularization: Semi-Supervised Setting;Measured (labeled) points: discriminant structure Unmeasured (unlabeled) points: geometrical structure;Measured (labeled) points: discriminant structure Unmeasured (unlabeled) points: geometrical structure;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Unlabeled Data to Estimate Geometry;Laplacian Regularized Least Square (Belkin and Niyogi, 2006);Active Learning;Objective: Guide the selection of the subset of data points that gives the most amount of information. Experimental design: select samples to label Manifold Regularized Experimental Design Share the same objective function as Laplacian Regularized Least Squares, simultaneously minimize the least square error on the measured samples and preserve the local geometrical structure of the data space. ; , In order to make the estimator as stable as possible, the size of the covariance matrix should be as small as possible. D-optimality: minimize the determinant of the covariance matrix ; Select the first data point such that