Abstract:Over the past few decades, dimensionality reduction has been widely exploited in computer vision and pattern analysis. This paper proposes a simple but effective nonlinear dimensionality reduction algorithm, named Maximal Linear Embedding (MLE). MLE learns a parametric mapping to recover a single global low-dimensional coordinate space and yields an isometric embedding for the manifold. Inspired by geometric intuition, we introduce a reasonable definition of locally linear patch, Maximal Linear Patch (MLP), which seeks to maximize the local neighborhood in which linearity holds. The input data are first decomposed into a collection of local linear models, each depicting an MLP. These local models are then aligned into a global coordinate space, which is achieved by applying MDS to some randomly selected landmarks. The proposed alignment method, called Landmarks-based Global Alignment (LGA), can efficiently produce a closed-form solution with no risk of local optima. It just involves some small-scale eigenvalue problems, while most previous aligning techniques employ time-consuming iterative optimization. Compared with traditional methods such as ISOMAP and LLE, our MLE yields an explicit modeling of the intrinsic variation modes of the observation data. Extensive experiments on both synthetic and real data indicate the effectivity and efficiency of the proposed algorithm.

An embedding dimension reduction algorithm based on sparse analysis

Local Linear Embedding in Dimensionality Reduction Based on Small World Principle

Local Linear Embedding with Adaptive Neighbors.

A new global embedding algorithm

Discriminative sparse embedding based on adaptive graph for dimension reduction

Nonlinear Dimensionality Reduction by Locally Linear Inlaying

Stable Local Dimensionality Reduction Approaches

Small World Neighborhood Optimized Local Linear Embedding Algorithm

Approximate Orthogonal Sparse Embedding for Dimensionality Reduction

Self-organized locally linear embedding for nonlinear dimensionality reduction

Local Discriminant Space Alignment

Local Neighborhood Embedding for Unsupervised Nonlinear Dimension Reduction.

Dimensionality Reduction for Text Using LLE

Towards mining the density and discriminative information of data in dimensionality reduction

Signed Laplacian Embedding for Supervised Dimension Reduction.

Dimensionality reduction by low-rank embedding

A modified local linear embedding algorithm based on neighbour selection.

Dimensionality Reduction By Using Sparse Reconstruction Embedding

Local Linear Transformation Embedding

Maximal Linear Embedding for Dimensionality Reduction

Sparse Local Discriminant Projections for Feature Extraction