Improvement on LPP Algorithm in Manifold Learning Leads to OLPP Algorithm

MATLAB 2.68M 250 views 0 downloads 1 credits
Tags:

Resource Overview

Enhancement of the Locality Preserving Projections (LPP) algorithm in manifold learning results in the Optimized Locality Preserving Projections (OLPP) algorithm, with applications in table recognition.

Detailed Documentation

In this document, we present an improvement on the Locality Preserving Projections (LPP) algorithm within the manifold learning framework, resulting in a novel algorithm termed Optimized Locality Preserving Projections (OLPP). The enhancement focuses on optimizing the projection matrix computation through iterative eigenvalue decomposition, which improves feature extraction efficiency and discrimination power. Key implementation steps include constructing the affinity matrix using k-nearest neighbors, computing the Laplacian matrix, and solving the generalized eigenvalue problem. Additionally, we conduct table recognition tasks by integrating OLPP-based feature reduction with convolutional neural networks for layout analysis. These advancements enable more effective pattern discovery and structural analysis in complex datasets, particularly for document processing applications.

You May Also Like